Heartbleed and static analysis

[Walter Bright]

On 4/11/2014 4:36 AM, Paulo Pinto wrote:
> This will never change as we (me and Walter) discussed on a parallel thread.
>
> The way arrays decay into pointers cannot be fixed while keeping backwards
> compatibility.
>
> Algol, PL/I and Mesa had bounds checked arrays, with the option to disable them
> if required, but C designers decided against it.
>
> The idea was that developers would use lint for such purposes, what very few do,
> even in 2014.
>
> I am convinced that this will only get fixed by a generation change.


The change I suggest is purely additive, meaning it will not break any existing 
code. If C can add VLAs, it can add this.

I have talked about this change with C people a few years ago, and I just kinda 
got a blank look in return. Maybe, though, C people will finally get fed up with 
these sorts of avoidable problems.