[gfortran,testsuite,committed] Do not xfail gfortran.dg/nint_2.f90 on ppc64-linux

Joseph S. Myers joseph@codesourcery.com
Fri Sep 14 14:56:00 GMT 2007


On Fri, 14 Sep 2007, François-Xavier Coudert wrote:

> Sorry. But actually, there is more to it.
>   -- ppc64-linux / glibc-2.4-20070330 passes the test, with both -m32
> and -m64 [1]
>   -- ppc-linux / glibc-2.6.1 fails, with both -m32 and -m64 [2]
>   -- ppc-linux / glibc-2.4-20060617 fails with -m32 (don't know about -m64) [3]
> 
> As glibc-2.6.1 was released on 20070731, I think the fact that
> ppc64-linux passes is not due to the bug being fixed in glibc, but to
> the target being ppc64 and not a ppc.

No, because there is in general *no difference* between a compiler 
generating code for N-bit GNU/Linux that is configured for one target and 
one configured for another target.  Both powerpc-linux and powerpc64-linux 
can and often do default to 32-bit, the only difference being whether the 
person configuring the compiler prefers to give it the "powerpc64" name.

Those particular compilers may have varying default CPUs, long double 
lengths, etc., and likewise glibc may have been configured with different 
options or compilers.  But it's the differences in how things were 
configured and built that are relevant, not the target triplets.

To work out the reasons you'd need to enquire further into the 
configuration of those compilers and of those glibcs and of the compilers 
used to build those glibc.  But from the results you give it looks like 
passing on ppc64 is a matter of luck and the XFAIL is appropriate on all 
powerpc*-*-linux* (without regard to multilib) unless and until the 
underlying bug is identified and fixed.

-- 
Joseph S. Myers
joseph@codesourcery.com


More information about the Gcc-patches mailing list