MinGW --mno-align-double

[Dimitrie O. Paun]

=======
From: Hans Leidekker 
To: [e-mail]

On Saturday 27 March 2004 02:20, Dmitry Timoshkov wrote:
> generated.c:542: Test failed: TYPE_ALIGNMENT(*(LPWIN32_STREAM_ID)0) == 8
> (expected 4) generated: 842 tests executed, 0 marked as todo, 1 failure.
> - NOTE: the same test compiled by cl.exe (MS compiler) works as expected,
> - so that's most likely a gcc bug. (gcc version 3.3.1 (cygming special))

Same here (GCC 3.2.3 (mingw special 20030504-1)). It appears 
that this compiler defaults to --malign-double. When I
specify --mno-align-double the test succeeds.
=======

Alexandre, can we add this flag to our makefiles? If so,
I can produce a patch...

[Alexandre Julliard]

"Dimitrie O. Paun"  writes:
> Same here (GCC 3.2.3 (mingw special 20030504-1)). It appears 
> that this compiler defaults to --malign-double. When I
> specify --mno-align-double the test succeeds.
> =======
>
> Alexandre, can we add this flag to our makefiles? If so,
> I can produce a patch...

No, I think it's better to avoid this. Our test environment is useful
for testing not only Wine but also Mingw, we have already found a
number of bugs that way. We would lose that possibility if we started
adding workarounds that hide the Mingw bugs.

[Hans Leidekker]

On Thursday 22 April 2004 19:37, Alexandre Julliard wrote:
> > Alexandre, can we add this flag to our makefiles? If so,
> > I can produce a patch...
>
> No, I think it's better to avoid this. Our test environment is useful
> for testing not only Wine but also Mingw, we have already found a
> number of bugs that way. We would lose that possibility if we started
> adding workarounds that hide the Mingw bugs.

Interestingly the diff for the MinGW compiler (against stock GCC) has
this comment:

"We also align doubles to 64-bits for MSVC default compatibility."

I will investigate this a bit more.

-Hans