Difference between parsing of void() and int()
Asked Answered
N

2

12

Having read about the most vexing parse, I experimented a bit and found this program. There are two very similar lines. One of them yields warnings in both g++7 and clang++-3.9, another does not.

int main() {
  void(); // no warning
  int(); // warning: statement has no effect
}

In the second line a default-constructed object of type int is created and immediately destroyed, thus unused. But what happens in the first line? If it was parsed the same way, it should be an error because it is illegal to create an object of type void. On the other hand, it does not look like a function declaration as well.

Noel answered 21/2, 2018 at 18:28 Comment(9)
Hmm, could it possibly be related to this? #34289344Raimund
clang ast -> docs -> standardEgwan
@Raindrop7 Try -Wall -Wextra, my results are with those.Noel
I knew I saw similar, possible dup of Returning a void?Egwan
@AndrewSun It is not a castEgwan
i guess void is not a data type. It is used to declare different things in arguments function a(void) , returns return void and, of course, universal pointer void *. But it is a completely different thing than int. So, no wonder that compiler does different treatment for them. Though i would prefer it gives some warning in the first place as well.Platitude
Casting to void is used to remove warning about unused variable. They probably treat this expression the same way.Sluff
@Sluff it is an explicit type conversion it is neeed so that templated functions do the right thing I don't know why people look at this like casting.Egwan
void() might also be used to avoid evil comma operator: (foo<Is>(), void(), 0)....Sluff
P
5

There is no difference in parsing. Both cases are covered by simple-type-specifier followed by optional parenthesized expression-list.

The semantic meaning is specified in C++17 (N4659) [expr.type.conv]/2:

If the type is cv void and the initializer is () , the expression is a prvalue of the specified type that performs no initialization. Otherwise, the expression is a prvalue of the specified type whose result object is direct-initialized with the initializer.

This specifically says that void() is a prvalue of type void.

Now, I'm sure that it is not intended that a prvalue of type void be illegal, as it is a common occurrence, e.g. (void)x; or calling a void function!

But I can't find where in the Standard it says that temporary materialization should be suppressed for void prvalues. The [class.temporary]/2 seems to say that a discarded-value expression always materializes a temporary; and it is an error to materialize a prvalue of incomplete type. Maybe it is a defect in the standard.


The difference in warning about "unused value" is probably because an unused value of type void is a common occurrence and it would not be helpful to warn about.

Phreno answered 28/3, 2018 at 3:13 Comment(0)
A
0

It is parsed the same way.

Warnings don't come from the parser. They arise during semantic analysis. The SA noticed that a value was created an destroyed by int(); without being read or written.

In the void case, there is no value, so no warning.

Adz answered 28/3, 2018 at 0:28 Comment(0)

© 2022 - 2024 — McMap. All rights reserved.