Home/ Questions/Q 8800071
In Process

The Archive Base Latest Questions

Editorial Team
  • 0
Asked: 2026-06-14T00:29:24+00:00

Suppose you have a C++ class Foo, and you say: Foo* foos = new

  • 0

Suppose you have a C++ class Foo, and you say:

Foo* foos = new Foo[SOME_CONSTANT];
memset(foos, 0, sizeof(Foo)*SOME_CONSTANT);
//or the bzero equivalent

and that Foo has a data member Bar* barPtr. Will the above operation guarantee that
barPtr will be NULL? (i.e. zero). I ran into a case in gdb where this didn’t hold for memset and I’m curious why.

I know the above is probably bad practice but I’m asking for curiosity.

I realized that I had an extra , i was doing sizeof(Foo)*SOME_CONSTANT in the memset…

Share

Leave an answer

You must login to add an answer.

Need An Account,

1 Answer

  1. Editorial Team

    On the one hand, it is implementation-defined. memset or bzero will fill the pointer value with all-zero bit pattern, which is not guaranteed to be a physical representation of null pointer on the given platform. It is not even guaranteed to produce a valid pointer value, meaning that you might end up with so called trap representation, which triggers undefined behavior when accessed.

    On the other hand, bzero is (or used to be) a part of POSIX specification. If memory serves, POSIX requires (or at least used to require) null-pointers to be represented by all-zero bit pattern, meaning that on POSIX systems it will indeed set a pointer to null.

    But again, the guarantee given by POSIX is just an example of an implementation-specific property mentioned in the first part of my answer. The language (C or C++) provides no such guarantees.

    • 0