Ruined Reputation: static const int...

[hendrixj]

I put my rep on the line was completely embarrassed.
First, I used a static const int in a function to size an array:

int function()
{
    static const int SIZE(10);
    char buf[SIZE];
    ...
}

This works on some versions of g++, but a guy running g++ (version 4. something) on a Lynux box had a error for it. The error indicated that char buf was being sized with a non-constant integral expression.
The compiler, for some reason, was not able to determine that SIZE should be replaced with 10.

Furthermore, C++ style teaches to use "static const int..." instead of #define. At least I was taught that.

I thought that not only did this enforce type safety, but any good compiler would optimize it away and you wouldn't waste space.

A guy here told me he put a "static const int ..." in a .h file and a separate copy of the const was created in each compilation unit that included the .h.

I think these are compiler errors. What say ye?

[Danny]

Actually, your reputation is untarnished -- something I can't say about your compiler, though.
The code is fine and should compile under any C++ compiler, including a C++98 compiler. I know that GNU have had some bugs with static const data, and I guess this bug is probably discussed on one of their forums (I'll have to check it) but for the time being, you can rest assure that this is perfectly valid C++ code. Try to replace the parentheses with a = sign. Perhaps this will convince the compiler that SIZE is a constant expression...

[hendrixj]

Actually, I should have put in the initial post that when I substituted = for the constructor form for initializing the constant, the Compiler was happy.

On the second issue, I found out that the guy used "no optimization" when compiling. No wonder the objects were not optimized away.

[Danny]

Even if the object weren't optimized away, they still make a const expression so the code should compiler in either case (i.e., optimized and unoptimized).