[geos-devel] test failures

[Sandro Santilli]

Theoretically -ffloat-store is already used when available,
is that working for you guys ?

There's also a ticket to replace -ffloat-store with
-fexcess-precision=standard
http://trac.osgeo.org/geos/ticket/504

--strk;

On Tue, Aug 20, 2013 at 10:06:54AM -0400, Greg Troxel wrote:
> 
> Mike Toews <mwtoews at gmail.com> writes:
> 
> > On 20 August 2013 21:30, Sandro Santilli <strk at keybit.net> wrote:
> >> Could you please open one ?
> >> The testsuite succeeds on my Ubuntu 64bit system.
> >
> > Funny, this problem appears to be only on a 32-bit OS, but not 64-bit.
> > Greg, is your NetBSD 32-bit?
> >
> > Issue opened:
> > http://trac.osgeo.org/geos/ticket/652
> 
> The system I tested on is "i386", which is one of many flavors of 32-bit
> CPUs.
> 
> On amd64 (x86_64), floating point tends to use SSE instructions, which
> do not have extended precision temporary values.   But on i386, floating
> point registers have more bits, and hence -ffloat-store.   So it would
> be interesting to see
> 
>   results on arm, sparc, sparc64 (and alpha and vax :-)
> 
>   results on i386 when compiling with -ffloat-store
> 
>   results with clang instead of gcc