Triage: PR7385

[Mark Mitchell <mark at codesourcery dot com>]

I'm going to downgrade some PRs that, while regressions, do not look
to me to be likely to be easily fixed and do not look to me to be
likely to have major impact.

This is a bit unilateral, and we usually try to avoid that, so I want
there to be feedback in the procedure.  I will send mail about the
downgrade, and if anyone objects we can undo the downgrade.

The first such PR is PR 7385.  

This is a C++ PR; the code in question is:

  typedef enum { v11=1, v12=2 } e1 __attribute__ ((packed));

We used to accept this code, now we do not; the root cause of that is
that we don't attached the packed attribute to the enum, but rather to
the typedef.  The problem is that we have already committed to the
size of e1 after seeing the closing curly brace for the enum; applying
the packed attribute here is too late.

The C front end does not manifest this problem in quite the same way,
but it does not handle it as the user expects either -- it does not
pack the enum, but instead silently ignores the attribute.  So, the
real bug here is that we do not issue an error message when
encountering the packed attribute that late in the game.

Therefore, I have downgraded the PR, and altered it to reflect that.

-- 
Mark Mitchell                   mark@codesourcery.com
CodeSourcery, LLC               http://www.codesourcery.com