Re: c++/9927: Compiler requires definition of static member constants

[Wolfgang Bangerth <bangerth@ticam.utexas.edu>]

The following reply was made to PR c++/9927; it has been noted by GNATS.

From: Wolfgang Bangerth <bangerth@ticam.utexas.edu>
To: Andrew Bell <acbell@iastate.edu>
Cc: gcc-bugs@gcc.gnu.org, <gcc-gnats@gcc.gnu.org>
Subject: Re: c++/9927: Compiler requires definition of static member  constants
Date: Mon, 3 Mar 2003 16:28:53 -0600 (CST)

 > >     The compiler is allowed to require that you provide a
 > >     definition of static member constants. I think, they are
 > >     even mandated. That's exactly what happens in your case.
 > >
 > >     You'll notice, however, that if you compile with optimization,
 > >     the compiler actually replaces references to these
 > >     variables by their values, which is probably what you
 > >     want.
 > 
 > Why, then, does the program compile and link if the "?:" construct is 
 > replaced with a "if (...) else (...)", even if the static const int's are 
 > not defined.
 
 By chance? The standard requires you to provide a definition of the 
 variable, so the compiler's free to make use of this requirement at its 
 own descretion. This might include certain contexts but not others.
 
 > I thought this was a special case for const int's.
 
 No, the special case is that you can give an initializer. But you can't 
 omit the definition.
 
 W.
 
 -------------------------------------------------------------------------
 Wolfgang Bangerth             email:            bangerth@ticam.utexas.edu
                               www: http://www.ticam.utexas.edu/~bangerth/