[Bug c++/20316] attribute((packed)) and call by reference

["falk at debian dot org" <gcc-bugzilla at gcc dot gnu dot org>]

------- Additional Comments From falk at debian dot org  2005-03-04 12:51 -------
(In reply to comment #0)

> Probably, this is not a bug but a feature

Indeed, it is.
> but what can I do
> to my program, which needs a lot of packed structures and runs fine
> with gcc3.3?

Nothing. It only "worked" by accident in 3.3, and will crash on many architectures.

> I understand, that my code may not work on cpus needing some specific
> alignment, but where is the problem with char?

It would be completely useless to make an exception for char here.

> And if this is a cpu issue,
> why do the pointers work, but _not_ the references?

Having pointers point to things that are not correctly aligned for their
type does *not* work. It might be easier to fool the compiler to not reject
it, though, and it might accidentally do what you meant it to do on some
architectures.


-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
         Resolution|                            |INVALID


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=20316