Re: Packing fixes (supersedes my previous packing fix patch)

[Dmitry Timoshkov]

"Hans Leidekker" <hans(a)it.vu.nl> wrote:
> This time I tried to be exhaustive. I used the following algorithm
> to find all structures that need 8 byte alignment:
>
> find all types typedef:ed to __int64
>   find all structures embedding any of the types found
>   REPEAT
>     find all types typedef:ed to any of the structures found
>     find all structures embedding any of the structures or types found
>   UNTIL no new types or structures are found
>
> This patch includes changes to .idl files as well as to the 
> corresponding .h files. Is that OK? Or should I remove changes 
> to .h files?

I think that a correct way deciding whether a struct is correctly
packed is to make sure that our generated tests for packing issues
work under MSVC, then just compare the results. Otherwise we risk
to miss something.

-- 
Dmitry.

Packing fixes (supersedes my previous packing fix patch)

[Hans Leidekker]

On Monday 03 May 2004 14:05, Dmitry Timoshkov wrote:
> I think that a correct way deciding whether a struct is correctly
> packed is to make sure that our generated tests for packing issues
> work under MSVC, then just compare the results. Otherwise we risk
> to miss something.

Sure, but until my patch to winapi_test and the regeneration
of the tests we weren't testing 8 byte alignments. Now we can,
so anyone who has a recent MSVC is invited to build and run the
generated tests ;-).

I see that Alexandre has put in a much better patch than mine 
that makes alignment default to 8 when it comes to 64 bit types,
which is what MSVC does.

 -Hans