This is the mail archive of the libc-alpha@sourceware.org mailing list for the glibc project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: double test failures on AMD machines


On 2014.01.30 at 07:25 +0000, Eric Wong wrote:
> Hey, I'm hitting some test failures on current master[1] on
> two of my AMD x86_64 machines, logs + /proc/cpuinfo here:
> 
> http://yhbt.net/w/test-ildoubl.out.gz
> http://yhbt.net/w/test-ldouble.out.gz
> http://yhbt.net/w/cpuinfo.gz
> 
> http://yhbt.net/s/test-ildoubl.out.gz
> http://yhbt.net/s/test-ldouble.out.gz
> http://yhbt.net/s/cpuinfo.gz
> 
> Both are Debian 7.0 x86_64 systems, built with
> gcc (Debian 4.7.2-5) 4.7.2
> Nothing non-default in ./configure besides --prefix=/tmp/g
> Copied libgcc_s.so.1 and libstdc++.so.6 over from the OS installation
> over to /tmp/g/lib
> 
> I couldn't reproduce these on an Intel Haswell machine.
> Was testing for [BZ #15132] avoid stat/fstat in statvfs/fstatvfs,
> but it happens without my patch (and my patch doesn't touch FP :)

That is expected. Please run "make regen-ulps" in your build directory.
Then copy the newulps file, e.g.:
"cp math/NewUlps ~/glibc/sysdeps/x86_64/fpu/libm-test-ulps".
And finally post the output of: 
"git diff sysdeps/x86_64/fpu/libm-test-ulps"
here.

-- 
Markus


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]