CWG3185 [decl.pre] static_assertion-declarations should not be name-declarations #894

Activity

1. YexuanXiao

   changed the title [-][decl.pre] static_assertion-declarations should not be named-declarations[/-] [+][decl.pre] static_assertion-declarations should not be name-declarations[/+] on Apr 28, 2026
2. jensmaurer commented on May 2, 2026

   

   jensmaurer

   on May 2, 2026

   Member

   More actions

   CWG3185
3. jensmaurer

   changed the title [-][decl.pre] static_assertion-declarations should not be name-declarations[/-] [+]CWG3185 [decl.pre] static_assertion-declarations should not be name-declarations[/+] on May 2, 2026
4. term-est commented on Jun 19, 2026

   

   term-est

   on Jun 19, 2026

   Member

   More actions

   linkage-specification: is in the form of:
   extern unevaluated-string name-declaration

   This has the side effect that:

   extern "C" consteval { };
   extern "C" static_assert(true);    
   extern "C" [[fumo]];
   extern "C" namespace reimu = std;
   extern "C" ;

   is well-formed according to our grammar, and no rule appears to make them ill-formed; the linkage semantics just have nothing meaningful to apply to.

   For consistency (and also because extern static_assert is equally bonkers as export static_assert), we should make these ill-formed as well.

   GCC and Clang are compliant, while MSVC and EDG rejects extern "C" [[attribute]]. externing a consteval block works on GCC 16.1 as it has reflection support.

   Proposed resolution for the export also makes export asm-declaration ill-formed, but I am not certain if extern asm-declaration should also be made ill-formed.

   Not sure whether this should be a new issue or considered as part of this one. Let me know if this needs to be its own issue and we don't consider the usecases above as NAD. If so I can open up a new issue for restricting the linkage-specification.

   👍React with 👍1Reacted by Yexuan Xiao