"The memory could not be written" error

**elmvin_s** · 12-12-2009

Hi guys,

I need some help here.

I executed my visual c++ program for several thousand iterations and at a point, my program halted and this error message appeared:

The instruction at "0x0040628e" referenced memory at "0x00000004". The memory could not be "written".

Does anyone know what went wrong?

Thank you.

**grumpy** · 12-12-2009

Your program performed an invalid operation with a pointer that had the result of attempting to write to an invalid area of memory. The error message is the way windows tells you it has detected and refused to allow the invalid operation.

If your code used arrays (rather than pointers) your code may have fallen off the end of an array. For example, if array x had 10 elements, your code attempted to modify x[n] where n >= 10.

There are many ways code can do this sort of thing: you will need to work out what the instruction is, and what line(s) of source code it corresponds to.

**EVOEx** · 12-12-2009

Originally Posted by grumpy

Your program performed an invalid operation with a pointer that had the result of attempting to write to an invalid area of memory. The error message is the way windows tells you it has detected and refused to allow the invalid operation.

If your code used arrays (rather than pointers) your code may have fallen off the end of an array. For example, if array x had 10 elements, your code attempted to modify x[n] where n >= 10.

There are many ways code can do this sort of thing: you will need to work out what the instruction is, and what line(s) of source code it corresponds to.

Actually, the memory address "0x00000004" indicates that it is likely that he tried to write to an element of a pointer that is uninitialized or set to NULL.

Your code is likely to look something like this:

Code:

int *p;
p = NULL; /* Might not be set, the compiler might have done this automatically */
...
p[1] = something;

It doesn't have to be the first index that is written to; if the type is a char it was probably the 4th element.

**elmvin_s** · 12-12-2009

Hi,

Thanks for the replies. Could be due to memory leakage as well?

**EVOEx** · 12-12-2009

Originally Posted by elmvin_s

Hi,

Thanks for the replies. Could be due to memory leakage as well?

... No.

**elmvin_s** · 12-12-2009

Dear EVOEx,

I am very new to C++. Could you further explain the problem?

The program could run fine for the first 2,000 iterations. It is only after that the error message appeared. If the problem was writting an element of a pointer which is uninitialised, wouldn't the program crash at the very beginning of execution, i.e. the first iteration?

Thanks for your help. Really appreciate it.

**iMalc** · 12-12-2009

It wouldn't be directly caused by a memory leak, but what might happen is that you might be using new no_throw, or even malloc (which most of the time shouldn't appear in C++ code) and the request might be failing and return NULL. Then your code is probably not checking for this error and is continuing on blindly with a NULL pointer.

Well that's one way at least. Do you know that you have a memory leak?

**grumpy** · 12-12-2009

Originally Posted by EVOEx

Actually, the memory address "0x00000004" indicates that it is likely that he tried to write to an element of a pointer that is uninitialized or set to NULL.

That is actually an invalid pointer operation, as I described in my previous post.

Originally Posted by EVOEx

Your code is likely to look something like this:

Code:

int *p;
p = NULL; /* Might not be set, the compiler might have done this automatically */
...
p[1] = something;

It doesn't have to be the first index that is written to; if the type is a char it was probably the 4th element.

That's only one of many possible scenarios by which code might write to an address with a value of 4. I've also seen such things come about through errors where the programmer thought he was doing this;

Code:

     int *p = some_valid_address();
     *p += result_of_long_series_of_computations();

but left out the asterix in the second line. While compilers often warn about such things (eg visual C++ complains about a suspicious pointer conversion, IIRC) programmers often ignore the warnings.

In a program that fails after many iterations, this type of scenario is more likely than code explicitly incrementing or molesting a NULL pointer.

However, there are many, many, other possibilities.

**jeffcobb** · 12-13-2009

Originally Posted by grumpy

That is actually an invalid pointer operation, as I described in my previous post.

That's only one of many possible scenarios by which code might write to an address with a value of 4. I've also seen such things come about through errors where the programmer thought he was doing this;

Code:

     int *p = some_valid_address();
     *p += result_of_long_series_of_computations();

but left out the asterix in the second line. While compilers often warn about such things (eg visual C++ complains about a suspicious pointer conversion, IIRC) programmers often ignore the warnings.

In a program that fails after many iterations, this type of scenario is more likely than code explicitly incrementing or molesting a NULL pointer.

However, there are many, many, other possibilities.

My money is on the second scenario or a derivation thereof where a ptr is being incremented and overflows to zero....

Jeff

**King Mir** · 12-13-2009

A debugger is good for spotting the cause of this sort of thing. Use it to find exactly where the error occurs while stepping through your code. Watch expressions and break points will help you.

Thread: "The memory could not be written" error

Thread Tools

Search Thread

Display

"The memory could not be written" error

Similar Threads

Quantum Random Bit Generator

Connecting to a mysql server and querying problem

Dikumud

Please Help - Problem with Compilers

ras.h errors