Strict aliasing and chains revisited

[Sebastian Huber]

On 10/27/2010 02:10 AM, Chris Johns wrote:
> On 25/10/10 11:49 PM, Sebastian Huber wrote:
>> On 10/26/2010 07:15 AM, Chris Johns wrote:
>>>
>>> Why the packed attribute ?
>>
>> In the thread from 2006
>>
>> http://www.rtems.com/ml/rtems-users/2006/november/msg00096.html
>>
>> some said that this my be necessary on some architectures to guarantee
>> the
>> overlapping.  I am not sure about this.  We may omit the packed
>> attribute and
>> see what the test suites say.
>>
> 
> I would rather we see a reason, discuss it and then decide what happens.
> Adding packed to anything makes it become very difficult to remove and
> know nothing breaks.

I removed the packed attribute and the test suite passes on ARM, SPARC and MIPS.

In the thread from 2006 the packed attribute was put into the discussion due to
some concerns with the ARM EABI.  I found nothing that indicates a 8 byte
structure alignment.  We only have a 8 byte stack alignment with the EABI.  I
run the test suite on ARM with my new EABI based tool chain and it works
without the packed attribute

-- 
Sebastian Huber, embedded brains GmbH

Address : Obere Lagerstr. 30, D-82178 Puchheim, Germany
Phone   : +49 89 18 90 80 79-6
Fax     : +49 89 18 90 80 79-9
E-Mail  : sebastian.huber at embedded-brains.de
PGP     : Public key available on request.

Diese Nachricht ist keine geschäftliche Mitteilung im Sinne des EHUG.