Clang should emit pedantic warnings for conversions to/from pointers to qualified arrays

[rjmccall]

Bugzilla Link	9082
Version	trunk
OS	All
CC	@ostash,@zygoloid

Extended Description

According to the C standard, conversions like the following are technically casting away const-ness, because array types are never qualified: qualifiers apply only to the element type.
const void * -> const int ()[5]
const int * -> const int ()[5]

Similarly, this is not supposed to be a valid qualification conversion:
int ()[10] -> const int ()[10]

These are kindof arbitrary restrictions that do make life unnecessarily difficult for real users; I think clang's model (where the array and element are simultaneously qualified) really makes a lot more sense. Nonetheless, we should at the very least provide -pedantic warnings about these conversions not being permitted by the standard(s).

[Endilll]

Clang 16 and GCC 13 don't warn about any of those conversions:
https://godbolt.org/z/YaEcne8aP

[llvmbot]

@llvm/issue-subscribers-c

[llvmbot]

@llvm/issue-subscribers-clang-frontend

[AaronBallman]

Current wording as of C2x is 6.7.3p10: If the specification of an array type includes any type qualifiers, both the array and the element type are so-qualified. ...

This differs from C17 where the array is not qualified but the element is. The changes came in from https://www.open-std.org/jtc1/sc22/wg14/www/docs/n2607.pdf.

I think it's reasonable to treat this paper as effectively being a DR instead of a new behavior in C2x. WDYT?

[Endilll]

Yeah, I don't think it's worth our scarce resources.

[rjmccall]

I agree with both Aaron's analysis and his conclusions.