[[maybe-unused]] vs __attribute__((unused)) vs (void) var for assert variables

[s-barannikov]

I don’t mind using it on functions. I think nobody would oppose that, because the alternative is LLVM_ATTRIBUTE_UNUSED.

When it comes to variables, this becomes more sensitive, because there are many more of them. There couldn’t be more low level component of a product than a coding style. Not only the code is already written, but the habbits are developed. The Big Switch can be carried out, but I can imagine people silently asking “But why? I’m so used to it.”.

@kparzysz-quic

@danilaml

The advantage of [[maybe_unused]] attribute is that it can apply to pretty much anything. I.e. you can mark functions/arguments with it, if there is some rarely used value (like for printf-like functions).

If we want consistency, then the attribute is the way to go.

There is no contradiction, we can consistently use it on functions. An analogy is the ‘const’ qualifier. We use it, but we’re not obsessed with it.

So far I’ve seen justifications for using the attribute, but not arguments for switching to it.
I won’t object if I see the use of the attribute in a review, but I don’t want to be forced to use it.
(The same applies to the other RFC that proposes switching to C++-style casts.)