Dimitry Andric writes:
> I am attaching a patch for qt4-corelib, which disables USE_GCC=any
> again, and adds a diff to fix up qsimd_p.h.
Thank you very much. Qt has been fixed upstream, and I've applied the
patch to qt4-corelib (and reverted all the uses of USE_GCC=any) in
r317896.
Dimitry Andric writes:
> On 2013-05-08 15:00, Raphael Kubo da Costa wrote:
>> Dimitry Andric writes:
>>> On 2013-05-08 13:18, Raphael Kubo da Costa wrote:
This still looks weird to me: QT_HAVE_3DNOW should only be defined if
the test in config.tests/unix/3dnow/3dnow.cpp passes. It's su
On 2013-05-08 16:11, Raphael Kubo da Costa wrote:
Dimitry Andric writes:
And it does similar includes for other intrinsics. Note that it tests
both QT_HAVE_SSE3 (which is a configure-time feature), and __SSE3__,
which is defined by the compiler, if it supports SSE3 in the current
mode. That i
Dimitry Andric writes:
> And it does similar includes for other intrinsics. Note that it tests
> both QT_HAVE_SSE3 (which is a configure-time feature), and __SSE3__,
> which is defined by the compiler, if it supports SSE3 in the current
> mode. That is, if you either pass -msse3 on the command
Dimitry Andric writes:
> On 2013-05-08 13:18, Raphael Kubo da Costa wrote:
>> This still looks weird to me: QT_HAVE_3DNOW should only be defined if
>> the test in config.tests/unix/3dnow/3dnow.cpp passes. It's supposed to
>> fail to build with clang-i486 just like qsimd.cpp.
>
> Those tests do no
On 2013-05-08 13:23, Raphael Kubo da Costa wrote:
Raphael Kubo da Costa writes:
This still looks weird to me: QT_HAVE_3DNOW should only be defined if
the test in config.tests/unix/3dnow/3dnow.cpp passes. It's supposed to
fail to build with clang-i486 just like qsimd.cpp.
[...]
I'm at $WORK ri
On 2013-05-08 13:18, Raphael Kubo da Costa wrote:
Dimitry Andric writes:
However, for 3DNow intrinsics, the header has something different:
#if defined(QT_HAVE_3DNOW)
#include
#endif
As you can see, there is no test whether the compiler defines __3dNOW__,
so if the configuration sta
Raphael Kubo da Costa writes:
> This still looks weird to me: QT_HAVE_3DNOW should only be defined if
> the test in config.tests/unix/3dnow/3dnow.cpp passes. It's supposed to
> fail to build with clang-i486 just like qsimd.cpp.
>
> [...]
>
> I'm at $WORK right now, and ref10-i386 lacks some of th
Dimitry Andric writes:
> However, for 3DNow intrinsics, the header has something different:
>
> #if defined(QT_HAVE_3DNOW)
> #include
> #endif
>
> As you can see, there is no test whether the compiler defines __3dNOW__,
> so if the configuration stage enables QT_HAVE_3DNOW (as it does for
Hi,
I finally had some time to figure out why devel/qt4-corelib could not be
built with clang. In particular, src/corelib/tools/qsimd.cpp fails
with:
clang++ -c -O2 -pipe -fno-strict-aliasing -pthread
-I/usr/local/include/glib-2.0 -I/usr/local/include -O2 -fvisibility=hidden
-fvisibility-inl
10 matches
Mail list logo
