On 11/03/2010 17:54, bjoern michaelsen - Sun Microsystems - Hamburg
Germany wrote:
> On Thu, 11 Mar 2010 17:26:12 +0100
> Herbert Duerr <[email protected]> wrote:
> 
>> That was my point. The rule that got us the hardcoded WIN16-style
>> code then is not IMHO is not a good guide for the future. If the
>> timeless native types had been used it would have grown with the
>> platform.
> 
> And thus would be ABI incompatible whenever the native types "grows".
> Naa, thats not better. This way we are, in theory, at least still free
> to make a sal_Int16 an 32-Bit value, whenever _we_ like to(*). With a
> native type, we are bound to whatever the platform deems appropriate.

presumably when the native type sizes change, the compiler runtime will be
ABI incompatible anyway, so that is a non-issue.

imho, fixed size types are mandatory only on I/O paths: network protocols,
file formats, and such.

note that due to remote UNO and bridges the UNO APIs effectively
constitute a network protocol, which is why the C++ binding uses fixed
size types.

> BR,
> 
> Bjoern
> 
> (*) Unless there is code that is _relying_ on overflows -- but that
> stuff breaks too with "native type growth".

-- 
"In other words, the answer to the question 'what is a medium/large
 application?' is simply this:  Any application that is so difficult to
 manage as a result of the proliferation of side-effects, that it often
 requires a committee of authoritarianism to approve a change that is
 almost certainly going to cause devastating consequences." -- Tony Morris


---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to