On 29/12/2024 09.33, Jₑₙₛ Gustedt wrote:
> on Sun, 29 Dec 2024 00:43:09 +0100 you (Jens Maurer
> <jens.maurer_at_[hidden]>) wrote:
>
>> On 28/12/2024 00.51, Jₑₙₛ Gustedt wrote:
>>> on Sat, 28 Dec 2024 00:26:41 +0100 you (Jens Maurer via Liaison
>>> <liaison_at_[hidden]>) wrote:
>>>
>>>> const char s1[] = "abc"; // #1
>>>> const char s2[] = ("abc"); // #2
>>>>
>>>> #1 is the canonical way to initialize a character array with
>>>> a string literal, and is supported by both C and C++.
>>>>
>>>> Recently, a question came up whether #2 is valid, too.
>>>> C++ uses grammar non-terminals in this area of the specification,
>>>> and it seems pretty clear that parenthesized string literals don't
>>>> satisfy the rules, and thus #2 is ill-formed.
>>>>
>>>> C23 superficially has the same rules as C++ in this area,
>>>> but I'd appreciate some confirmation.
>>>
>>> The corresponding phrase only refers to string literals themselves
>>> (the syntax element) and not expressions, as would be a string
>>> literal enclosed in parenthesis, 6.7.11 p7, first sentence:
>>
>> Ok, so we both have the same understanding of the C23 rules.
>>
>>> The initializer for an array shall be either a string literal,
>>> optionally enclosed in braces, or a brace-enclosed list of
>>> initializers for the elements.
>>>
>>>> (Quite a few implementations appear to accept #2.
>>>
>>> do they ? I never would have suspected this to work.
>>
>> Yes, unanimously so (as far as C++-related implementations
>> are concerned):
>>
>> https://godbolt.org/z/EPTrn1PdG
>
> It seems that it only appeared recently in clang (18 or so), which I
> find surprising.

Selecting the strict standards conformance mode yields an error for
EDG; the rest of the compilers continue to accept without a warning.

Clang13 and clang5 also accept without a warning when using "-xc -std=c11 -pedantic",
so I'm not following the statement "appeared recently in clang".

> Perhaps people refer to
>
> A parenthesized expression is a primary expression. Its type, value,
> and semantics are identical to those of the unparenthesized
> expression.
>
> But this widening says nothing about syntax so it does not apply, here.

Agreed.

Jens