From: jameskuyper@alumni.caltech.edu   
      
   On 2026-01-13 11:11, Andrey Tarasevich wrote:   
   > On Mon 1/12/2026 9:36 AM, Michael S wrote:   
   >> But I was interested in the "opinion" of C Standard rather than of gcc   
   >> compiler.   
   >> Is it full nasal UB or merely "implementation-defined behavior"?   
   >   
   > ... And, of course, it is as   
   > "implementation-defined" as any other UB in a sense that the standard   
   > permits implementations to _extend_ the language in any way they please,   
   > as long as they don't forget to issue diagnostics when diagnostics are   
   > required (by the standard).   
      
   "implementation defined" is a term defined by the standard. It does not   
   carry it's ordinary English definition "defined by the implementation".   
   Instead, it means "unspecified behavior where each implementation   
   documents how the choice is made". Unless the standard explicity says   
   that the behavior is implementation-defined, the fact that any   
   particular implementation chooses to define it is irrelevant.   
      
   >>> Perhaps there's a switch in GCC that would outlaw the classic "struct   
   >>> hack"... But in any case, it is not prohibited by default for   
   >>> compatibility with pre-C99 code.   
      
   The struct hack has always technically had undefined behavior, but in   
   practice almost all C90 implementations allowed it to work. In C99   
   flexible array members were added, which allows the struct hack to work   
   with fully-defined behavior using slightly different syntax. The struct   
   hack itself is still just as undefined as it ever was, and because of   
   the invention of flexible array members, is increasingly likely to not   
   be supported.   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)