Re: list of "show-stoppers"?

[Jeffrey A Law <law at cygnus dot com>]

  In message <19990702072336K.mitchell@codesourcery.com>you write:
  > >>>>> "Jason" == Jason Merrill <jason@cygnus.com> writes:
  > 
  >     Jason> I prefer my patch below, which I checked into the mainline
  >     Jason> a few days ago.  I think that the current behavior is much
  >     Jason> more useful than what your patch does; if the user's code
  >     Jason> is nonconformant, we should add a pedwarn, not just
  >     Jason> gratuitiously break it.
  > 
  > I didn't understand that your patch was designed to fix this problem;
  > the tree I was working in didn't have your patch.
  > 
  > But, the user's code *is* conformant; it just doesn't mean what the
  > user thought.
  > 
  >   typedef const struct { int i; } S;
  > 
  > is perfectly legal; it just doesn't give the struct the name `S' for
  > linkage purposes.  As you can see from the test-case attached, doing
  > this prevents us from correctly warning about subsequent things like:
  > 
  >   void f (S);
  > 
  > in which a type without linkage is used in an external function type.
Where do we stand with this issue?  I'm not qualified to evaluate which
of the two patches is best for the release.

jeff