[Icc-avr] Packing of variables in union

[Richard]

What John says, actually, the things that will break the code would 
be either a) the endianness change, or b) the alignment changes. 
Neither is likely but I wouldn't want to promise it will work forever 
and ever. Who knows, the next group of AVRs may be 
super-turbo-internet-enabled-AVRs that would benefit from word aligment...

At 03:00 PM 5/18/2007, John Baraclough wrote:
>Hi Johan,
>
>Sorry, but that is very bad programming practice.
>
>K&R states: "It is the responsibility of the programmer to keep 
>track of what type is currently stored in a union; the results are 
>machine dependent if something is stored as one type and extracted as another."
>
>Thus writing:
>
>u.w16 = 123;
>c = u.b8;
>
>may not return the expected value of 123 into 'c'. Furthermore, it 
>is unsafe to assume that what works in one release will also work in 
>the next. I'm sure that Richard will try his utmost to retain 
>reverse compatibility, but K&R does not require it.
>
>HTH.
>
>All the best for now,
>John
>
>
>At 20:47 18/05/2007, you wrote:
>>Hi,
>>
>>can I trust ICCAVR to *always* put all members of a union at
>>the same starting address?
>>
>>Example:
>>   given the following declaration:
>>
>>union
>>{
>>   short w16;
>>   char b8;
>>   long l32;
>>   char abFourBytes[4];
>>   char abSevenBytes[7];
>>} u;
>>
>>is this true:
>>
>>   b8 == (w16 & 0xFF) == (l32 & 0xFF) == abFourBytes[0] == abSevenBytes[0] ?
>>
>>I have in some dull memory that alignment in struct and union is allowed
>>to be "impementation dependent" but I can see no reason use other than
>>8-bit alignment (== no "padding") in the AVR (?).
>>
>>I don't have my trusty old K&R at hand right now so I can't check 
>>the rules...
>>
>>
>>Regards
>>Johan Bodin, SM6LKM
>>
>>_______________________________________________
>>Icc-avr mailing list
>>Icc-avr at imagecraft.com
>>http://dragonsgate.net/mailman/listinfo/icc-avr
>
>
>_______________________________________________
>Icc-avr mailing list
>Icc-avr at imagecraft.com
>http://dragonsgate.net/mailman/listinfo/icc-avr

// richard (This email is for mailing lists. To reach me directly, 
please use richard at imagecraft.com)