Architectures/PowerPC/WorkQueue

Work Queue

This is the central point where work on build failures is organized, you can find here a list of failed builds, list of actual tasks and who works on them.

Failed builds

• failed builds between 2014-06-23 and 2014-07-04 = http://fedora.danny.cz/ppc/20140704-ppc-failed.txt

Tracker bugs

• ppc/ppc64 = https://bugzilla.redhat.com/show_bug.cgi?id=1071880
• ppc64le = https://bugzilla.redhat.com/show_bug.cgi?id=1051573
• alpha blocker = https://bugzilla.redhat.com/show_bug.cgi?id=1140359
• alpha freeze exception = https://bugzilla.redhat.com/show_bug.cgi?id=1043122
• beta blocker = https://bugzilla.redhat.com/show_bug.cgi?id=1146697
• final blocker = https://bugzilla.redhat.com/show_bug.cgi?id=1146698

Tasks

• higher on list mean higher prio

package	problem	who	bugzilla	notes	resolved in build
llvm	v3.5 doesn't like gcc 4.9	davids	1123103 1133508	Trying to solve the issue in glibc headers
hornetq				needs a bootstrap build and regular build afterwards
mdadm	FTBFS due bad format in printf		1125883		mdadm-3.3.1-6.fc21
mariadb	tests failing on BE & LE ???	mtoman	1096787	latest version built, but one of the tests fails occasionally
R	3.1.1 doesn't build on LE	davids	1136388	(R coredumps)
coreutils/findutils/gettext/augeas/...	hanging test-lock from gnulib test-suite	N/A		isolated test case at http://jcajka.fedorapeople.org/gnulib-test-small.tar.gz
systemd	missing ppc64le support in ver >= 215	N/A		http://lists.freedesktop.org/archives/systemd-devel/2014-July/020884.html	systemd-215-10.fc21
dyninst	FTBFS on ppc64le	sharkcz	1113991		dyninst-8.1.2-10.fc21
gperftools	missing ppc64le support	sharkcz		update 2.2.1	gperftools-2.2.1-1.fc21
ceph	OOM due too many CPUs and too few memory	sharkcz		run build on builder with only 4 or 6 CPUs online - added RFE for redhat-rpm-config
dbus-python	test failing			test fails on ppc64 only with latest buildroot	dbus-python-1.2.0-5.fc21 rebuilt with pygobject3-3.13.3-1.fc21 in buildroot
libdb	FTBFS with openjdk8 - http://ppc.koji.fedoraproject.org/koji/taskinfo?taskID=1941862	jcajka	1119258	fails on all archs	libdb-5.3.28-7.fc21
python3-cairo	failing tests - http://ppc.koji.fedoraproject.org/koji/taskinfo?taskID=1942017	jcajka		probably the same problem was with pycairo	python3-cairo-1.10.0-10.fc21
atlas	missing ppc64le support	michel_mno	1080073	still pending discussion upstream	temporary build atlas-3.10.2-0.fc21.dh.1 until rebased by maintainer
erlang-rebar erlang-getopt erlang-erlydtl erlang-neotoma	missing bootstrap	michel_mno	1122088	I am not authorised to commit erlang-rebar	all those packages have been bootstrapped and built
python	FTBFS in primary - http://koji.fedoraproject.org/koji/taskinfo?taskID=7135864 - the maintainer is looking on it => root cause are changes in glibc and openssl		glibc 1119769	we need a new build because python-2.7.7-2.fc21 uses the old libffi ABI on LE	python-2.7.7-3.fc21 built
strace	"failed to trace fstatat/fstatat64 properly" - http://ppc.koji.fedoraproject.org/koji/taskinfo?taskID=1960025	sharkcz	1122323		strace-4.8-5.fc21
cxxtools	failing test "xmlserializer::testDouble: EXCEPTION"	menantea	1108800	patch proposed
xorg-x11-apps and busybox	struct winsize not defined	hamzy	1122714	should be fixed in glibc 2.20

Detailed plans

python (DONE)

• http://ppc.koji.fedoraproject.org/koji/taskinfo?taskID=1962223 which uses latest gcc-4.9.1-2.fc21 and glibc-2.19.90-30.fc21, but still fails on setegid test
• plan agreed with Siddhesh
  • (DONE as glibc-2.19.90-31.fc22) he builds new glibc in rawhide and assures there is no problem in primary (i686), no builds in f21 until we are sure it works on all platforms
  • (DONE) we rebuild this one in f21 as scratch build - http://ppc.koji.fedoraproject.org/koji/taskinfo?taskID=1964230
  • then rebuild python with it, mean a local mock build, on both ppc64 and ppc64le and wait for the result
    • if all OK, I'll forward the info to Siddhesh
    • if BAD, then we will work on isolating the failing test and give him access to the machine
    • isolating the test could mean rebuild without test suite to have the python binaries that can be then used to execute the test

Various notes

• recently I have seen a stuck guile process (during build of graphviz-2.38.0-8.fc21 on ppc-le-builder2), very likely due the threading issue (waiting in futex), after attaching strace to the hung process, the process got unblocked