Re: [xslt] 1.1.15 test failure on Redhat 7.1, RHEL 2.1/x86, 3/x86, 4/x86



On Fri, Nov 11, 2005 at 10:34:03PM -0600, Albert Chin wrote:
> I've encountered the same test suite failure on Redhat 9, RHEL 3/x86,
> and 4/x86:
>   ## Running exslt common tests
>   add-duration.1 result
>   43c43 < result : P1000Y11M31D --- > result : P1000Y11M31DT 9.32232069317251e-11S
>   gmake[5]: Leaving directory `/opt/build/libxslt-1.1.15/tests/exslt/date'
> 
> $ cd /opt/build/libxslt-1.1.15/tests/exslt/date
> $ diff -u add-duration.1.out add-duration.1.res 
> --- add-duration.1.out  2002-10-15 07:28:04.000000000 -0500
> +++ add-duration.1.res  2005-11-11 21:35:33.739708788 -0600
> @@ -40,7 +40,7 @@
>  add-duration : PT23H59M0S + PT61S
>  result       : P1DT1S
>  add-duration : P1000Y11M30DT23H59M0.1S + PT59.9S
> -result       : P1000Y11M31D
> +result       : P1000Y11M31DT 9.32232069317251e-11S
>  add-duration : P1000Y11M30DT23H59M1S + PT59S
>  result       : P1000Y11M31D
>  add-duration : P1000Y11M30D + -P1DT9S
> 
> Any ideas? BTW, I don't have this problem on Redhat 7.1, RHEL 2.1/x86,
> 3/AMD64, 4/AMD64. Kinda odd.
> 
> Ok, never mind. Compiler bug. Building with -O0 fixed it. The bug is
> in date.c. Build it alone with -O0 fixes things.

Thanks to Eric Botcazou, this is GCC bug PR323. However, the bug won't
be fixed :) A workaround is to build date.c with -O2 -ffloat-store.

-- 
albert chin (china thewrittenword com)


[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]