Re: FreeBSD_HEAD_amd64_gcc - Build #1340 - Still Failing

> On Jul 5, 2016, at 11:54, Ngie Cooper (yaneurabeya) <yaneurabeya_at_gmail.com> wrote:
> 
>> 
>> On Jul 5, 2016, at 11:52, Dimitry Andric <dim_at_FreeBSD.org> wrote:
>> 
>> On 05 Jul 2016, at 18:03, jenkins-admin_at_freebsd.org wrote:
>>> 
>>> FreeBSD_HEAD_amd64_gcc - Build #1340 - Still Failing:
>>> 
>>> Build information: https://jenkins.FreeBSD.org/job/FreeBSD_HEAD_amd64_gcc/1340/
>>> Full change log: https://jenkins.FreeBSD.org/job/FreeBSD_HEAD_amd64_gcc/1340/changes
>>> Full build log: https://jenkins.FreeBSD.org/job/FreeBSD_HEAD_amd64_gcc/1340/console
>> ...
>>> /builds/FreeBSD_HEAD_amd64_gcc/usr.sbin/bhyve/rfb.c: In function 'sse42_supported':
>>> /builds/FreeBSD_HEAD_amd64_gcc/usr.sbin/bhyve/rfb.c:885:17: error: 'bit_SSE42' undeclared (first use in this function)
>>> return ((ecx & bit_SSE42) != 0);
>>>               ^
>> 
>> So, this is because clang's and gcc's versions of cpuid.h use slightly different naming for these bits:
>> 
>> clang's cpuid.h:
>> 
>> #define bit_SSE41       0x00080000
>> #define bit_SSE42       0x00100000
>> 
>> gcc's cpuid.h:
>> 
>> #define bit_SSE4_1      (1 << 19)
>> #define bit_SSE4_2      (1 << 20)
>> 
>> Unfortunately there are more bit defines that differ.  No standardization on this point, it seems. :-/
>> 
>> For this specific compile error, we could put in a little crutch like:
>> 
>> #if defined(bit_SSE4_2) && !defined(bit_SSE42)
>> #define bit_SSE42 bit_SSE4_2
>> #endif
>> 
>> Thoughts?
> 
> 	Seems ok to me. I was going to submit a patch to fix the other issues with bhyve (because I am getting annoyed by the build failure emails).

Actually, there’s a better #define:

/usr/include/x86/specialreg.h:#define   CPUID2_SSE41    0x00080000

Thanks,
-Ngie