Pointers stored in the list getting corrupted??

[bmsce engg]

Hi,
Have been struggling to find the reason for this occasional problem with pointers that I am storing in a list (STL C++).
My class name is : Tamper

All I do is this:

1- Instantiate Tamper with new and push the address in a list of type Tamper. The number of pushes are huge, may be in hundreds (dont think it matters as long as i have space).

Tamper *ptr=new Tamper();
list<Tamper *> = mylist;
mylist.push_bac k(ptr);

2- At a later point, I simply use list iterators and retrieve the stored pointers and call "delete" on them, which sounds straight forward and it works fine for me. But once in a while, I see that my process dumps core with SIGSEGV or SIGBUS, and stack points to "delete" operation.

list<Tamper *>::iterator it = mylist.begin();

for(; it != mylist.end() ; ++it)
{
delete *it;
*it=0;
}

How do I get rid of this occasional but consistent problem while freeing the memory. I don't think I am trying to release someone else's memory.

Appreciate your help here.

[weaknessforcats]

What you show is straightforward . You delete pointers you allocated and you set those deleted pointers to zero to avoid an exception by deleting them again.

Are you sure you never pushed a local pointer? That wil do it every time.

Is this a multi-threaded program? If so, the delete and he subsequent setting of the pointer to zero is no thread safe.

[bmsce engg]

Thanks for your reply first of all!!!
This is not multithreaded, and there is no pushing of local pointer. I already did something like this just to make sure I am deleting my own pointers:

for(; it != mylist.end() ; ++it)
{
if ( typeid(*it) == typeid(Tamper *) )
{
delete *it;
*it=0;
}
}


This works in a similar way as before. But dumps core as usual.

The point here is, somehow the addresses are getting corrupted.
- Is there a better way to validate the address before deleting?
- Could there be environment issue here?

[weaknessforcats]

I ran your code for 1000 pointers and failed to produce a crash.

There's something else going on that's not in the code you have shown. Using naked pointers in C++ is very dangerous. The Tamper* should be a handle object.

There is no way to validate an address. The most you can do is say that a zero address is invalid. All non-zero addresses are presumed valid.

Is there by chance a delete of the Tamper* somewhere else in the code and at that place the pointer is not set to zero?

[mac11]

I think you'd be better off having your list hold smart pointers (such as boost::shared_p tr). Then you don't have to explicitly delete anything and are much less likely to run into problems with raw pointers.

For an example see:

http://www.boost.org/doc/libs/1_46_1/libs/smart_ptr/example/shared_ptr_example.cpp

[bmsce engg]

Thanks for your answer.
I will check and make sure I am not deleting this address somewhere else, though I am sure I am not doing it.
I will come back to you weaknessforcats after some more check.

[bmsce engg]

Thanks mac11... I will check the possibility of using boost shared ptrs.