If you shouldn't throw exceptions in a destructor, how do you handle errors in it?

Most people say never throw an exception out of a destructor - doing so results in undefined behavior. Stroustrup makes the point that "the vector destructor explicitly invokes the destructor for every element. This implies that if an element destructor throws, the vector destruction fails... There is really no good way to protect against exceptions thrown from destructors, so the library makes no guarantees if an element destructor throws" (from Appendix E3.2).

This article seems to say otherwise - that throwing destructors are more or less okay.

So my question is this - if throwing from a destructor results in undefined behavior, how do you handle errors that occur during a destructor?

If an error occurs during a cleanup operation, do you just ignore it? If it is an error that can potentially be handled up the stack but not right in the destructor, doesn't it make sense to throw an exception out of the destructor?

Obviously these kinds of errors are rare, but possible.

Throwing an exception out of a destructor is dangerous.
If another exception is already propagating the application will terminate.

#include <iostream>

class Bad
{
    public:
        // Added the noexcept(false) so the code keeps its original meaning.
        // Post C++11 destructors are by default `noexcept(true)` and
        // this will (by default) call terminate if an exception is
        // escapes the destructor.
        //
        // But this example is designed to show that terminate is called
        // if two exceptions are propagating at the same time.
        ~Bad() noexcept(false)
        {
            throw 1;
        }
};
class Bad2
{
    public:
        ~Bad2()
        {
            throw 1;
        }
};


int main(int argc, char* argv[])
{
    try
    {
        Bad   bad;
    }
    catch(...)
    {
        std::cout << "Print This\n";
    }

    try
    {
        if (argc > 3)
        {
            Bad   bad; // This destructor will throw an exception that escapes (see above)
            throw 2;   // But having two exceptions propagating at the
                       // same time causes terminate to be called.
        }
        else
        {
            Bad2  bad; // The exception in this destructor will
                       // cause terminate to be called.
        }
    }
    catch(...)
    {
        std::cout << "Never print this\n";
    }

}

This basically boils down to:

Anything dangerous (i.e. that could throw an exception) should be done via public methods (not necessarily directly). The user of your class can then potentially handle these situations by using the public methods and catching any potential exceptions.

The destructor will then finish off the object by calling these methods (if the user did not do so explicitly), but any exceptions throw are caught and dropped (after attempting to fix the problem).

So in effect you pass the responsibility onto the user. If the user is in a position to correct exceptions they will manually call the appropriate functions and processes any errors. If the user of the object is not worried (as the object will be destroyed) then the destructor is left to take care of business.

An example:

std::fstream

The close() method can potentially throw an exception. The destructor calls close() if the file has been opened but makes sure that any exceptions do not propagate out of the destructor.

So if the user of a file object wants to do special handling for problems associated to closing the file they will manually call close() and handle any exceptions. If on the other hand they do not care then the destructor will be left to handle the situation.

Scott Myers has an excellent article about the subject in his book "Effective C++"

Edit:

Apparently also in "More Effective C++"
Item 11: Prevent exceptions from leaving destructors

Throwing out of a destructor can result in a crash, because this destructor might be called as part of "Stack unwinding". Stack unwinding is a procedure which takes place when an exception is thrown. In this procedure, all the objects that were pushed into the stack since the "try" and until the exception was thrown, will be terminated -> their destructors will be called. And during this procedure, another exception throw is not allowed, because it's not possible to handle two exceptions at a time, thus, this will provoke a call to abort(), the program will crash and the control will return to the OS.

We have to differentiate here instead of blindly following general advice for specific cases.

Note that the following ignores the issue of containers of objects and what to do in the face of multiple d'tors of objects inside containers. (And it can be ignored partially, as some objects are just no good fit to put into a container.)

The whole problem becomes easier to think about when we split classes in two types. A class dtor can have two different responsibilities:

• (R) release semantics (aka free that memory)
• (C) commit semantics (aka flush file to disk)

If we view the question this way, then I think that it can be argued that (R) semantics should never cause an exception from a dtor as there is a) nothing we can do about it and b) many free-resource operations do not even provide for error checking, e.g. void free(void* p);.

Objects with (C) semantics, like a file object that needs to successfully flush it's data or a ("scope guarded") database connection that does a commit in the dtor are of a different kind: We can do something about the error (on the application level) and we really should not continue as if nothing happened.

If we follow the RAII route and allow for objects that have (C) semantics in their d'tors I think we then also have to allow for the odd case where such d'tors can throw. It follows that you should not put such objects into containers and it also follows that the program can still terminate() if a commit-dtor throws while another exception is active.

With regard to error handling (Commit / Rollback semantics) and exceptions, there is a good talk by one Andrei Alexandrescu: Error Handling in C++ / Declarative Control Flow (held at NDC 2014)

In the details, he explains how the Folly library implements an UncaughtExceptionCounter for their ScopeGuard tooling.

(I should note that others also had similar ideas.)

While the talk doesn't focus on throwing from a d'tor, it shows a tool that can be used today to get rid of the problems with when to throw from a d'tor.

In the future, there may be a std feature for this, see N3614, and a discussion about it.

Upd '17: The C++17 std feature for this is std::uncaught_exceptions afaikt. I'll quickly quote the cppref article:

Notes

An example where int-returning uncaught_exceptions is used is ... ... first creates a guard object and records the number of uncaught exceptions in its constructor. The output is performed by the guard object's destructor unless foo() throws (in which case the number of uncaught exceptions in the destructor is greater than what the constructor observed)

The real question to ask yourself about throwing from a destructor is "What can the caller do with this?" Is there actually anything useful you can do with the exception, that would offset the dangers created by throwing from a destructor?

If I destroy a Foo object, and the Foo destructor tosses out an exception, what I can reasonably do with it? I can log it, or I can ignore it. That's all. I can't "fix" it, because the Foo object is already gone. Best case, I log the exception and continue as if nothing happened (or terminate the program). Is that really worth potentially causing undefined behavior by throwing from a destructor?

Its dangerous, but it also doesn't make sense from a readability/code understandability standpoint.

What you have to ask is in this situation

int foo()
{
   Object o;
   // As foo exits, o's destructor is called
}

What should catch the exception? Should the caller of foo? Or should foo handle it? Why should the caller of foo care about some object internal to foo? There might be a way the language defines this to make sense, but its going to be unreadable and difficult to understand.

More importantly, where does the memory for Object go? Where does the memory the object owned go? Is it still allocated (ostensibly because the destructor failed)? Consider also the object was in stack space, so its obviously gone regardless.

Then consider this case

class Object
{ 
   Object2 obj2;
   Object3* obj3;
   virtual ~Object()
   {
       // What should happen when this fails? How would I actually destroy this?
       delete obj3;

       // obj 2 fails to destruct when it goes out of scope, now what!?!?
       // should the exception propogate? 
   } 
};

When the delete of obj3 fails, how do I actually delete in a way that is guaranteed to not fail? Its my memory dammit!

Now consider in the first code snippet Object goes away automatically because its on the stack while Object3 is on the heap. Since the pointer to Object3 is gone, you're kind of SOL. You have a memory leak.

Now one safe way to do things is the following

class Socket
{
    virtual ~Socket()
    {
      try 
      {
           Close();
      }
      catch (...) 
      {
          // Why did close fail? make sure it *really* does close here
      }
    } 

};

Also see this FAQ