[Buildroot] [autobuild.buildroot.net] Build results for 2013-11-22

Thomas De Schampheleire patrickdepinguin at gmail.com
Mon Nov 25 09:25:13 UTC 2013


On Sat, Nov 23, 2013 at 9:17 PM, Baruch Siach <baruch at tkos.co.il> wrote:
> Hi Thomas,
>
> On Sat, Nov 23, 2013 at 02:11:43PM +0100, Thomas De Schampheleire wrote:
>> On Sat, Nov 23, 2013 at 8:30 AM, Thomas Petazzoni
>> <thomas.petazzoni at free-electrons.com> wrote:
>> > Build statistics for 2013-11-22
>> > ===============================
>> >
>> >         success : 68
>> >        failures : 34
>> >        timeouts : 0
>> >           TOTAL : 102
>> >
>> [..]
>>
>> >       sh4a |                   poco-1.4.6p1 | NOK | http://autobuild.buildroot.net/results/8dc9bca5969d455cc30206fc1ae6e1d3ba6eb6ca/
>>
>> Error is:
>>
>> In file included from src/FPEnvironment_C99.cpp:37:0,
>>                  from src/FPEnvironment.cpp:48:
>> include/Poco/FPEnvironment_C99.h:56:30: error: 'FE_DOWNWARD' was not
>> declared in this scope
>> include/Poco/FPEnvironment_C99.h:57:30: error: 'FE_UPWARD' was not
>> declared in this scope
>> make[2]: *** [/home/test/test/2/output/build/poco-1.4.6p1/Foundation/obj/Linux/sh4a/release_shared/FPEnvironment.o]
>> Error 1
>> make[2]: Leaving directory
>> `/home/test/test/2/output/build/poco-1.4.6p1/Foundation'
>> make[1]: *** [Foundation-libexec] Error 2
>> make[1]: Leaving directory `/home/test/test/2/output/build/poco-1.4.6p1'
>> make: *** [/home/test/test/2/output/build/poco-1.4.6p1/.stamp_built] Error 2
>>
>> The same build failure has been discussed a long time ago here:
>> http://lists.busybox.net/pipermail/buildroot/2011-September/046013.html
>> when the poco package was introduced. Then, the fpenvironment was
>> disabled when the C library is uclibc.
>> In the current failure, the C library seems to be glibc.
>> I'm including Baruch on this one, who was involved in the above
>> thread. Baruch: any idea?
>
> I see the following comment in glibc sysdeps/sh/sh4/fpu/bits/fenv.h:
>
> /* The SH FPU supports two of the four defined rounding modes: round to nearest
>    and round to zero.  We use again the bit positions in the FPU control word
>    as the values for the appropriate macros.  */
>
> I don't know enough about fenv to know what the right solution here would be.
> The easiest one is to add --no-fpenvironment for sh4a. Is this acceptable?

I couldn't tell you. We already pass this option in one situation, at
first sight it would be acceptable to pass it in another.

Best regards,
Thomas



More information about the buildroot mailing list