Re: Subtleties of __attribute__((packed))

[Jan Blunck <jblunck@suse.de>]

On Wed, Dec 06, Russell King wrote:

> On Wed, Dec 06, 2006 at 04:54:39PM +0100, Jan Blunck wrote:
> > Maybe the arm backend is somehow broken. AFAIK (and I verfied it on S390 and
> > i386) the alignment shouldn't change.
> 

Once again: I refered to "packed attribute on the struct vs. packed attribute
on each member of the struct". The alignment shouldn't be different.

> Please read the info pages:
> 
> `packed'
>      This attribute, attached to an `enum', `struct', or `union' type
>      definition, specifies that the minimum required memory be used to
>      represent the type.
> 
>      Specifying this attribute for `struct' and `union' types is
>      equivalent to specifying the `packed' attribute on each of the
>      structure or union members.  Specifying the `-fshort-enums' flag
>      on the line is equivalent to specifying the `packed' attribute on
>      all `enum' definitions.
> 
> Note that it says *nothing* about alignment.  It says "minimum required
> memory be used to represent the type." which implies that the internals
> of the structure are packed together as tightly as possible.
> 
> It does not say "and as such the struct may be aligned to any alignment".
> 

And this is why it makes sense to think about align attribute when you use
packed.