`bugprone-unintended-char-ostream-output` reported with sufficient cast

[firewave]

#include <cstdint>
#include <iostream>

enum E : std::uint8_t {
    E1
};

void f(E e)
{
    std::cout << static_cast<unsigned char>(e);
}

<source>:10:15: warning: 'unsigned char' passed to 'operator<<' outputs as character instead of integer. cast to 'unsigned int' to print numeric value or cast to 'char' to print as character [bugprone-unintended-char-ostream-output]
   10 |     std::cout << static_cast<unsigned char>(e);
      |               ^  ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
      |                  static_cast<unsigned int>(static_cast<unsigned char>(e))

https://godbolt.org/z/vzPoa9jnd

CC @HerrCai0907

[llvmbot]

@llvm/issue-subscribers-clang-tidy

Author: Oliver Stöneberg (firewave)

```cpp #include <cstdint> #include <iostream>

enum E : std::uint8_t {
E1
};

void f(E e)
{
std::cout << static_cast<unsigned char>(e);
}

<source>:10:15: warning: 'unsigned char' passed to 'operator<<' outputs as character instead of integer. cast to 'unsigned int' to print numeric value or cast to 'char' to print as character [bugprone-unintended-char-ostream-output]
10 | std::cout << static_cast<unsigned char>(e);
| ^ ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
| static_cast<unsigned int>(static_cast<unsigned char>(e))

https://godbolt.org/z/vzPoa9jnd

CC @<!-- -->HerrCai0907 
</details>

[HerrCai0907]

I think it should be warned by default. since in this case, output are still invalid ascii char.
But I will add option to make check accept explicit cast.

[carlosgalvezp]

@firewave can you ellaborate on what you think is the problem? The cast does not solve the problem that the check is helping with, why should it not warn? Also, if you already have a cast, why not cast to the right type instead?

[firewave]

That cast does what it is supposed to do - enforce the usage of the ASCII value: https://godbolt.org/z/o746bxKMK

I chose that particular type since I did not want to cast from unsigned to signed.

[carlosgalvezp]

In that case the correct cast is to "char". There are 3 types of char, signed/unsigned char are numeric, and plain char is for text. Allowing such a cast would make sense as the intention of the author is clear.

[firewave]

There are 3 types of char, signed/unsigned char are numeric, and plain char is for text.

What?

Update (posted too soon): std::cout treats it as text no matter the signedness.

[carlosgalvezp]

What?

https://eel.is/c++draft/basic#fundamental-7

Type char is a distinct type that has an implementation-defined choice of “signed char” or “unsigned char” as its underlying type

Also:
https://stackoverflow.com/questions/16503373/why-is-char-distinct-from-signed-char-and-unsigned-char

signed char and unsigned char are numeric types, i.e. intended to be used in arithmetic computations. Plain char is not a numeric type, but instead intended to represent a text character.

I did not want to cast from unsigned to signed.

What problem do you get by doing so? It seems std::basic_ostream handles all 3 equally:
https://en.cppreference.com/w/cpp/io/basic_ostream/operator_ltlt2

[carlosgalvezp]

The docs also clearly state that in order to clearly express intent one should cast to char:

Or cast to char to explicitly indicate that output should be a character.

[firewave]

Ugh, more homework...

I did not want to cast from unsigned to signed.

What problem do you get by doing so?

I did not want to unnecessarily cast to a different signedness (the codebase in questions is really bad with that and I did not want to add to the pile of stuff I need to review at some point in the future).

[HerrCai0907]

If you want an integer output here, then cast to uint32 is better. if you want a char, then cast to char is more suitable. actually there are no signed and unsigned in ascii level.