C++

Forum

 
 
operator new[] and data initialization

yman (11)
Hi all!
I'm having a trouble with the operator new[] used to allocate an array of long unsigned int.
I rely on the operator that all the array items are initialized to zero, as mentioned in http://www.cplusplus.com/reference/std/new/operator%20new%5B%5D/ :
... it then automatically initializes or constructs each of the objects in the array (if needed). ...

Somewhere in my program I use the array items as conditions of an if clause. Nevertheless valgrind throws an error here saying "Conditional jump or move depends on uninitialised value(s)".

Here's a simplified code of current, problematic state:
1
2
3
 
unsigned long int *plData = new unsigned long int[width*height];
...
if(plData[0]) ...


I found out how to simply solve the problem:
1
2
3
4
 
unsigned long int *plData = new unsigned long int[width*height];
memset((void*)plData, 0, width*height*sizeof(unsigned long int));
...
if(plData[0]) ...


Yes, the solution is simple, but my question is: Isn't the memset() call redundant? I do think so since the data is declared to be initialized after the new[] call. Isn't the problem in valgrind? Or is it unsafe to rely on this functionality of the operator new[].

Now the program is compiled using g++, nevertheless it's also cross-compiled using arm-none-linux-gnueabi-g++ and possibly others, so I need to be sure it's not dependent on the compiler.

Thanks in advance!
yman
Last edited on
kbw (9488)
The memset is not redundant. Calling new in this way (on a POD array) does not initialise the memory block.
yman (11)
Well, this is really a completely new information to me. Good to know.
Thanks very much, kbw!
Topic archived. No new replies allowed.