Fix Issue 15286 - is(typeof(symbol))

[RazvanN7]

No description provided.

[dkorpel]

"semantically correct" is currently a poorly defined term. Is cast(ubyte*)0xDEADBEEF semantically correct inside a @safe function? The spec for __traits(compiles, X) also uses "semantically correct" but I recall there being subtle differences between that and is(typeof(X)).

I suppose it's out of scope for this issue though.

[ntrel]

Something like "to check whether an expression type checks correctly" would be more accurate. Errors not related to type checking are not done here AIUI. Could also suggest __traits(compiles) for full semantic checks.

[RazvanN7]

Semantically correct is essentially the same thing type checking from a compilers perspective. However, type-checking is a compiler developer specific term, whereas "semantically correct" is much more friendly to the user.

[ntrel]

@RazvanN7

Semantically correct is essentially the same thing type checking from a compilers perspective

That is not true when the context affects whether it compiles:

int i;
void f() pure
{
    pragma(msg, is(typeof(i++))); // true
    pragma(msg, __traits(compiles, i++)); // false
}

So the new docs are misleading that is(typeof(expr)) tests whether expr is semantically correct. Some semantic checks are skipped.

Also Don Clugston said in 2012:

It checks if the expression has a type. That's all.

https://forum.dlang.org/post/k6o04g$259s$1@digitalmars.com