Re: ntdll: Protect RtlAllocateHeap and RtlReAllocateHeap against integer overflows with large values of size.

2007-05-18 Thread Chris Robinson 
On Friday 18 May 2007 04:01:19 am Robert Shearman wrote:
 +    ULONGLONG llret = (ULONGLONG)a + b;
 +    if ((sizeof(SIZE_T)  sizeof(ULONGLONG))  (llret  0x))
 +        return FALSE;

WOuldn't this be more correct (as well as function when sizeof(SIZE_T) = 
sizeof(ULONGLONG)):

SIZE_T res = a + b;
return (res = a);

Re: ntdll: Protect RtlAllocateHeap and RtlReAllocateHeap against integer overflows with large values of size.

2007-05-18 Thread Robert Shearman 

Chris Robinson wrote:

On Friday 18 May 2007 04:01:19 am Robert Shearman wrote:
  

+ULONGLONG llret = (ULONGLONG)a + b;
+if ((sizeof(SIZE_T)  sizeof(ULONGLONG))  (llret  0x))
+return FALSE;



WOuldn't this be more correct (as well as function when sizeof(SIZE_T) = 
sizeof(ULONGLONG)):


SIZE_T res = a + b;
return (res = a);
  


An example that would break using your logic:
2 + (-1)

--
Rob Shearman

Re: ntdll: Protect RtlAllocateHeap and RtlReAllocateHeap against integer overflows with large values of size.

2007-05-18 Thread Chris Robinson 
On Friday 18 May 2007 05:12:30 am you wrote:
 An example that would break using your logic:
 2 + (-1)

SIZE_T (if it follows standard size_t) is unsigned, though. Adding a negative 
wouldn't be possible.

Re: ntdll: Protect RtlAllocateHeap and RtlReAllocateHeap against integer overflows with large values of size.

2007-05-18 Thread Robert Shearman 

Chris Robinson wrote:

On Friday 18 May 2007 05:12:30 am you wrote:
  

An example that would break using your logic:
2 + (-1)



SIZE_T (if it follows standard size_t) is unsigned, though. Adding a negative 
wouldn't be possible.


Yes, you're right. The second parameter should probably be SSIZE_T.

--
Rob Shearman