Re: Buffer overflow in psql

Tom Lane wrote:

> Ah so, that explains how come it noticed.  BTW, I see that somebody
> already changed the array size to 16 bytes in HEAD --- so it's just
> the back branches that need fixing.

Um, is that really considered a fix???  We all know that there's no  
guarantee at all, even in ANSI C, that unsigned int isn't bigger than  
32 bits, right?  There are still some weird architectures out there.

Whenever I need to print some integer x, I use code like this:

   char buf[1 + sizeof(x) * CHAR_BIT / 3]

I let the compiler figure out the length needed to print in octal,  
and use that as a (slight) over-estimate of the length for decimal.   
As a bonus, the type of x can be changed without having to track down  
this kind of crap.

Alternatively, the code in question could just cast to one of the  
newer fixed-length int types, like int32_t, although that has its own  
problems.

Sorry for the pedantry ...

- John D. Burger
   MITRE