Overflows in Phobos

[Steven Schveighoffer via Digitalmars-d]

On 7/26/16 2:44 PM, Timon Gehr wrote:
> On 26.07.2016 17:44, Johan Engelen wrote:
>> The compiler can assume it is unreachable code, but it has to keep it,
>
> That makes no sense. Those two statements are mutually exclusive.

I thought that assert(0) means that the compiler does not need to check 
any validity of code after the assert. I'd reword Johan's statement to 
say that the compiler can assume the programmer thought this was 
unreachable, so it can just crash. It can't compile the crash out, or 
then the program is in an invalid state!

This is not the user saying "I guarantee we won't go over the cliff", 
it's the user saying "If we get to this point, I have lost all 
guarantees of not going off the cliff, so shut off the engine". The 
compiler can make optimization assumptions for the code that *doesn't* 
take the branch, but it still needs the branch to make sure.

-Steve