Re: [usrmarshal] Adjust new tests to pass on win9x platforms.

[Jacek Caban]

Hi Jeremy,

On 12/17/09 12:13 AM, Jeremy White wrote:
> dlls/oleaut32/tests/usrmarshal.c |    9 ++++++---
>   1 files changed, 6 insertions(+), 3 deletions(-)
>
> diff --git a/dlls/oleaut32/tests/usrmarshal.c b/dlls/oleaut32/tests/usrmarshal.c
> index e9c53e9..1c770de 100644
> --- a/dlls/oleaut32/tests/usrmarshal.c
> +++ b/dlls/oleaut32/tests/usrmarshal.c
> @@ -306,7 +306,8 @@ static void test_marshal_LPSAFEARRAY(void)
>       buffer = HeapAlloc(GetProcessHeap(), 0, size);
>       init_user_marshal_cb(&umcb,&stub_msg,&rpc_msg, buffer, size, MSHCTX_DIFFERENTMACHINE);
>       next = LPSAFEARRAY_UserMarshal(&umcb.Flags, buffer,&lpsa);
> - ok(next - buffer == expected, "Marshaled %u bytes, expected %u\n", (ULONG) (next - buffer), expected);
> - ok(next - buffer == expected || next - buffer + sizeof(DWORD) == expected /*win9x*/,
> -
>        "Marshaled %u bytes, expected %u\n", (ULONG) (next - buffer), expected);

It looks like a good place to use broken().

Thanks,
     Jacek

[usrmarshal] Adjust new tests to pass on win9x platforms.

[Jeremy White]

> It looks like a good place to use broken().

I don't think it's broken on win98; it looks as though
they do 4 byte alignment prior to the data structure in win9x, 
and winnt and on seem to do 8 byte alignment prior to the
data chunk.  That results in a 4 byte difference.

Cheers,

Jeremy

[usrmarshal] Adjust new tests to pass on win9x platforms.

[Alexandre Julliard]

Jeremy White  writes:
> > It looks like a good place to use broken().
>
> I don't think it's broken on win98; it looks as though
> they do 4 byte alignment prior to the data structure in win9x, 
> and winnt and on seem to do 8 byte alignment prior to the
> data chunk.  That results in a 4 byte difference.

It should really do 8 byte alignment for 64-bit types. I think that
qualifies as broken.