enabled and fixed bugprone-narrowing-conversions clang-tidy warnings

[firewave]

No description provided.

[danmar]

I don't really agree here. ch is a unsigned char. all the bits will be saved in prev.

[firewave]

This does not change any behavior. It just makes the implicit conversions explicit (prev is char and currentToken is basic_string<char>).

[danmar]

I do know it doesn't change the behavior.

Casts are also bugprone and silence compilers. can bugprone-narrowing-conversions be solved without casts?

[firewave]

Casts are also bugprone and silence compilers. can bugprone-narrowing-conversions be solved without casts?

No, not if the narrowing is intentional. The point is to make the code explicit.

[danmar]

No, not if the narrowing is intentional. The point is to make the code explicit.

I feel that casts are dangerous because they hide all compiler warnings about all conversion errors. It's the sloppy approach to silencing warnings. Similar to writing --suppress=*. And there is no mechanism to detect redundant casts.

It does not say specifically that it tries to prevent warnings about the narrowing conversion. You could suggest a new cppcheck checker that will warn about this which provides a more explicit mechanism to hide the warnings. And I could approve that.

A similar code example where there is no warning:

void foo(unsigned int x) {
    char y = static_cast<char>(x);  // <- we have loss of precision here
}

Your cast hides warning about the sign conversion and it would also hide future warnings about loss of precision if we will have that bug. Real bugs can be hidden.

[danmar]

clang-tidy does not tell me there is likely mistake here.

int y;
int foo(int x) {
    y = static_cast<char>(x);
}

if I could indicate that my intention with the cast is only sign-conversion :

int y;
int foo(int x) {
    y = static_cast_sign_conversion<char>(x);
}

then we could warn here, the cast does not do what was intended.