How to deal with variables where the size differs for 32 and 64bit

[Paul Vriens]

Hi,

Just found the fix for the oledb32:convert tests on 64bit. We need to 
adjust our oledb.idl (see patch attached).

How should I deal with the ok() messages where we currently specify the 
32bit variant? I mean this will of course introduce warnings on the 
64bit side.

Do I need to create a wrapper for the ok() messages, with a _WIN64 check 
somewhere? Or use a different printf format identifier?

The same is also true for the TRACE() messages in our implementation.

[André Hentschel]

Am 05.12.2010 21:05, schrieb Paul Vriens:
> Hi,
>
> Just found the fix for the oledb32:convert tests on 64bit. We need to
> adjust our oledb.idl (see patch attached).

wow, good job!!!
> How should I deal with the ok() messages where we currently specify the
> 32bit variant? I mean this will of course introduce warnings on the
> 64bit side.

sry, actually i don't know, just wanted to thank you.

[Reece Dunn]

On 5 December 2010 20:05, Paul Vriens  wrote:
> Hi,
>
> Just found the fix for the oledb32:convert tests on 64bit. We need to adjust
> our oledb.idl (see patch attached).

Nice find.

Shouldn't these be using the _PTR types instead (ULONG_PTR, LONG_PTR
and friends?) -- then you don't need the 32/64-bit switch.
> How should I deal with the ok() messages where we currently specify the
> 32bit variant? I mean this will of course introduce warnings on the 64bit
> side.
>
> Do I need to create a wrapper for the ok() messages, with a _WIN64 check
> somewhere? Or use a different printf format identifier?

Does "%p" (pointer) work? Are there any other tests that handle 32/64-bit types?

- Reece

[Austin English]

On Sun, Dec 5, 2010 at 12:05 PM, Paul Vriens  wrote:
> Hi,
>
> Just found the fix for the oledb32:convert tests on 64bit. We need to adjust
> our oledb.idl (see patch attached).
>
> How should I deal with the ok() messages where we currently specify the
> 32bit variant? I mean this will of course introduce warnings on the 64bit
> side.
>
> Do I need to create a wrapper for the ok() messages, with a _WIN64 check
> somewhere? Or use a different printf format identifier?
>
> The same is also true for the TRACE() messages in our implementation.

Why not use an #ifdef there as well? A few other tests do it, for example:
http://source.winehq.org/git/wine.git/?a=blob;f=dlls/mshtml/tests/dom.c#l370...
http://source.winehq.org/git/wine.git/?a=blob;f=dlls/kernel32/tests/process....