ARM compiler misbehaves ?

[Richard Earnshaw]

On Thu, 2004-04-29 at 15:17, Grant Edwards wrote:
> On Thu, Apr 29, 2004 at 10:44:14AM +0100, Richard Earnshaw wrote:
> 
> > > >test_pointer now points to an odd address. Dereferencing it
> > > >will almost certainly not work right, since the compiler
> > > >doesn't generate run-time checks for misaligned pointers.
> > 
> > Wrong.
> 
> OK, which are you claiming?
> 
>  a) test_pointer points to an even address.
> 
>  b) Dereferencing the pointer works right.
> 
> or 
> 
>  c) The compiler generates run-time checks for misaligned
>     pointers.
> 

None of them.  I was referring to the statement:

> Actually, I'm pretty sure it is guaranteed to work as long as the 
> compiler can see the "packed" attribute during compilation (if it 
> couldn't, that would be a serious structural problem in your source
code!).

Which claimed it should work.  It doesn't, and isn't intended to.


> Can you provide any evidence for any of the three?  I know for
> a fact that none of the three claims is true for 2.9x and 3.x
> versions of gcc targetted for ARM or H8 architectures.  Which
> target architecture are you talking about?
>     
> > You can't take the address of a packed element and store it in a
> > normal pointer.
> 
> Last time I checked (with the ARM and H8 gcc's) you could. It's
> just that dereferencing the pointer won't work right. Copying
> to/from the value of the pointer with memcpy() works fine.
> 

No, even that isn't guaranteed, since the compiler is entitled to "know"
the implementation of memcpy, and therefore to implement it inline.

If you assign a pointer to a 'short *' and then try to dereference it,
then the compiler is entitled to assume that the value really is a legal
short * and has all the properties required of such a pointer.


> > Normal pointers can't handle unaligned objects.
> 
> What do you mean by "handle"?

Dereference.

R.

------
Want more information?  See the CrossGCC FAQ, http://www.objsw.com/CrossGCC/
Want to unsubscribe? Send a note to crossgcc-unsubscribe@sources.redhat.com