The patch has been posted at http://gcc.gnu.org/ml/gcc-patches/2010-11/msg00489.html and commited, thanks Jason for fixing this! (if this is a doublepost, sorry for that) Zdenek # > ---------------------------------------- # > On 11/07/2010 05:29 AM, Benoit Jacob wrote: # > > This code: # > > # > >>>>> static const unsigned x = (unsigned)-1; # > >>>>> # > >>>>> enum e { # > >>>>> ey = (int)x # > >>>>> }; # > > # > > relies on the constant global variable x being a true compile-time # > > constant, right? # > > But as far as I can see, it is impossible to guarantee that it # > > actually is, since in theory you could do # > > # > > const_cast<int*>(&x) = 0; # > # > The standard says, # > # > Except that any class member declared mutable (7.1.1) can be modified, # > any attempt to modify a const object during its lifetime (3.8) results # > in undefined behavior. # > # > So the above assignment has undefined behavior, which means the compiler # > is allowed to assume that it doesn't happen. # > # > > I guess that the only way is to say that the compiler remembers the # > > initializer value -1 for x and uses it instead of x itself. In other # > > words x itself is never used. # > # > Right. # > # > > I guess that this is where the "static" # > > matters here since inside of one translation unit the initializer # > > value -1 can be remembered? # > # > 'static' doesn't matter, just that it's const, integral, and initialized # > with a constant expression. # > # > Jason # > # > # > #
