C++ g++.dg/parse/parse3.C test

[Matt Austern]

We've got a test to verify that the parser accepts the following 
construct:
enum numbers { zero, one, two, three } __attribute__ ((packed)) ;

Trouble is, we're accepting this construct and ignoring it.  That's a 
bug.  (http://gcc.gnu.org/bugzilla/show_bug.cgi?id=14124) The fact that 
parse3.C passes is misleading, and it's only passing because the test 
isn't thorough enough.

I don't have a fix for c++/14124. and I don't have any plans to work on 
it.  What I do have is a patch that causes the compiler to emit a 
warning when attributes get ignored; that warning is useful for other 
reasons.  That patch causes parse3.C to start failing.  I'm not sure 
what the proper protocol is for cases like this.  Options include:
  1. Accept the fact that a test case will start failing when it 
previously succeeded, and decide that it's a good thing because it 
increases the visibility of a known bug.
  2. XFAIL parse3.C.
  3. Remove parse3.C altogether.

Suggestions?

			--Matt