Building packages with buildd and qemu-kvm

[Svante Signell]

Hi,

looking at the build log of the for a long time build-attempted lapack
from  bach.debian.net and my own build trial, my build progressed a much
longer in the tests than on bach. However, both builds ended strangely:

qemu-kvm: (during linking)
gfortran   schkrfp.o sdrvrfp.o sdrvrf1.o sdrvrf2.o sdrvrf3.o sdrvrf4.o
serrrfp.o slatb4.o slarhs.o sget04.o spot01.o spot03.o spot02.o chkxer.o
xerbla.o alaerh.o aladhd.o alahd.o alasvm.o  \
        ../../tmglib_LINUX.a ../../lapack_LINUX.a -lblas -o xlintstrfs
&& mv xlintstrfs ../xlintstrfs
make[3]: Leaving directory
`/home/srs/DEBs/lapack/lapack-3.3.0/testing/LIN'
Testing REAL LAPACK RFP protoype linear equation routines
../debian/trun 1800 ./xlintstrfs < stest_rfp.in > stest_rfp.out 2>&1
make[1]: *** [lapack_testing] Error 141
make[1]: Leaving directory `/home/srs/DEBs/lapack/lapack-3.3.0'
make: *** [testing/xeigtsts] Error 2
/usr/bin/fakeroot: line 1: kill: (4152) - No such process
dpkg-buildpackage: error: fakeroot debian/rules binary gave error exit
status 2

bach: (during execution of a test)
./xeigtsts exited with status 0 T
../debian/trun: line 39: 16217 Terminated              while true; do
    touch foo1 -t $(date +%Y%m%d%H%M.%S -d "$T seconds ago"); [ foo -ot foo1 ] 
&& kill -15 $j > /dev/null 2>&1 && break; sleep 1;
done
SBB: Testing banded Singular Value Decomposition routines
../debian/trun 1800 ./xeigtsts < sbb.in > sbb.out 2>&1
libfakeroot: read: Connection reset by peer
make[1]: *** [lapack_testing] Error 141
make: *** [testing/xeigtsts] Error 2
kill: 1: No such process

dpkg-buildpackage: error: /usr/bin/fakeroot debian/rules binary-arch gave error 
exit status 2
==> testing/sbb.out <==
 SBB:  NRHS =   2
 All tests for SBB passed the threshold ( 3000 tests run)
 End of tests
 Total time used =         0.06 seconds
make[1]: Leaving directory 
`/build/buildd-lapack_3.3.0-4-hurd-i386-vg7oTp/lapack-3.3.0'

Both tests end with a signal being sent for a kill from fakeroot in the
first case and in libfakreoot during a read operation. Both had happen
during redirection of stdin and stdout. Why issuing a kill in the
middle of execution? Are these functions also handled by libpipe? If
that is the case we might have a common denominator with the ghc6 build
problems I have. I've also seen very strange things happen not related to the
actual builds. A hurd/gnumach/libc bug??

BTW: Compiling atlas on kvm hangs when trying to find the L1 cache
size, but the buildd passes that stage. Are all buildds on xen, or is
some box running only hurd?