a896bd18   
   From: francis.glassborow@btinternet.com   
      
   On 05/09/2011 23:20, Javier Jimenez wrote:   
   > Hello   
   >   
   > The example code bellow compiles in gcc, and produces a core dump.   
   > When I noticed the error ( int&  i = i; ), I would expected two   
   > possible compiler behaviours:   
   >   - it does not compile.   
      
   Perfectly reasonable expectation and many years ago (about 1996) I   
   suggested that   
   int i = i;   
      
   self initialisation, be deemed an error. However there is an obscure   
   program technique that relies on this declaration being valid for a user   
   defined type and so WG21 were reluctant to remove it. However good tools   
   will identify it and warn you.   
      
   Now I know that what you wrote is not the same but I falls into a   
   similar parsing problem. Once i has been declared it will hide any outer   
   declaration of i. However I think that there is no reason why this form   
   of declaration should not be marked as ill-formed.   
      
   >   - the reference points to the "i" out of the "if context" (const int   
   > i)   
      
   Yes, that would be reasonable but it would also be inconsistent so I   
   would prefer it to be ill-formed.   
   >   
   > But my surprise is that it does compile... and crashes in runtime.   
   >   
   > Why?   
      
   Why does it compile? Probably because the compiler does not identify it   
   as erroneous That is probably because the C++ Standard does not   
   explicitly outlaw it though I suspect that it is actually ill-formed   
   implicitly because the reference does not refer to storage.   
      
   int i;   
   int & j = i;   
   int & k = j;   
      
   is, I think, well formed and valid.  We would need to explicitly require   
   that a reference was initialised to actual storage (though possibly in a   
   register) even if the compiler has to see through a number of   
   intermediate references. Actually wording that is not simple and, I   
   think, has to be pulled out as a special case. I.e.   
      
   type t& = t;   
   and   
   type t&(t);   
      
   would have to be made ill formed.   
      
   Try posting your example to comp.std.c++   
      
      
   --   
         [ See http://www.gotw.ca/resources/clcm.htm for info about ]   
         [ comp.lang.c++.moderated.    First time posters: Do this! ]   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)