Printable View
Hi all,
The following is not producing an error under VC++ 2005 Express.
Under Dev-C++, I get what I'm expecting. A compile-time error about the fact I'm trying to create an unitialized array of const integers.Code:const int *p = new const int[10]; // ();
Do you know what compiler option(s) should I set under VC++ 2005 in order for it to correctly flag this as an error?
maybe Dev-C++ compiler is wrong??
>>Do you know what compiler option(s) should I set under VC++ 2005 in order for it to correctly flag this as an error?
Probably none -- no compiler is 100% compilant with c++ standards.
I'm sorry... but this is confusing me...
I tried the following after I read a related article on MSDN :
The above also works on Dev-C++ provided I initialize p. But what is confusing me is that I'm apparently allowed to change the referenced value of a pointer to const.Code:int val = 12;
const int *p = new const int; //on purpose. *p is undefined here on VC++ 2005
int *temp = &val;
p = temp;
std::cout << *p << std::endl; //outputs 12!
True, I cannot if I simply tried *p = 12. But can someone please help me understand why can I do this?
Forget it. I found the answer... I'm not changing the contents of the pointer when I do p = temp. I'm changing the pointer itself. It is still following the rules for pointers to constant objects.
However, I'm still curious...
Even if undefined under VC++ 2005, wouldn't it make sense to have it at least issue a warning? After all, a uninitialized pointer to const is rather useless and seems (I think) a rather easy thing to catch.
Maybe Dev-C++ is wrong. I couldn't find yet information on the ISO or Stroustrup's C++. But my guts tell me MinGW is closer to the truth...
It might be a warning that they turn off, but allow you to turn on.Quote:
Even if undefined under VC++ 2005, wouldn't it make sense to have it at least issue a warning?
list of the warnings that are off by default:
http://msdn2.microsoft.com/en-US/library/23k5d385.aspx
Big list... Thanks Skorman. I thought that having warnings set at level 4 would be enough. Apparently not.
I've set "/Wall" and it was a nightmare :)
Lots of warnings. All related with the standard library. But not one about this issue.
So, apparently VC++ 2005 treats uninitialized const objects as undefined and issues no warning. Maybe this is the way it should be as Ancient Dragon suggested.
VC++ 2005 has its quirks but overall is much better than previous generations of MS compilers.