Both safe and wrong?

[Olivier FAURE]

On Thursday, 7 February 2019 at 01:04:44 UTC, Walter Bright wrote:
> I'm not seeing a problem with it.

Let me put this another way:

If you want to make sure only @safe code is ever run, you can 
slap @safe on your main function, and the program will only 
compile if every single other function is either @safe or 
@trusted.

However, @system variable declarations aren't infectious so far 
(and making them infectious would be a breaking change), which 
means the following code:

     import std.stdio;

     immutable int x = 1;
     int* y = cast(int*)&x;

     void main() @safe
     {
         *y = 2;
         writeln(x);
         writeln(*(&x));
         writeln(*y);
     }

compiles even though main(), a @safe function, ends up doing 
something unsafe (mutating a value declared as immutable).

One way to fix this would be to forbid using @system global 
variables in @safe functions, but this would definitely be a 
breaking change, unless global variable safety is determined by 
the compiler by default (which is its own can of worms).