Mailinglist Archive: opensuse-factory (331 mails)

< Previous Next >
[opensuse-factory] openSUSE reproducible builds status 2018-11
Hi,

last month's status:
https://lists.opensuse.org/opensuse-factory/2018-10/msg00242.html

I uploaded http://rb.zq1.de/compare.factory-20181121/ today
and rbstats are:
total-packages: 11068
build-tried: 11067
build-failed: 42
build-n-a: 104
build-succeeded: 10921
build-official-failed+na: 84
build-compare-failed: 299
build-compare-succeeded: 10622
bit-by-bit-identical: 10252
not-bit-by-bit-identical: 667


Last months' r-b project updates (including mine):
https://reproducible-builds.org/blog/posts/186/
https://reproducible-builds.org/blog/posts/185/
https://reproducible-builds.org/blog/posts/184/
https://reproducible-builds.org/blog/posts/183/
https://reproducible-builds.org/blog/posts/182/


Fixed core packages since last month:
gnome-builder https://build.opensuse.org/request/show/644025

grep https://build.opensuse.org/request/show/647618

infinipath-psm https://build.opensuse.org/request/show/644077

kdoctools fixed in https://bugs.kde.org/show_bug.cgi?id=398263

open-iscsi https://build.opensuse.org/request/show/644084

qpid-proton https://build.opensuse.org/request/show/644081

thai-fonts https://build.opensuse.org/request/show/626785


Nearly fixed:

evolution https://gitlab.gnome.org/GNOME/evolution/issues/196

xen https://build.opensuse.org/request/show/644624 - gets
(unavoidable?) variations from pesign-obs-integration.
One xen/binutils patch is hard to upstream


The reason why there are less fully reproducible packages today, is that
many .a files differed because I did these builds without this pending
patch: https://github.com/openSUSE/brp-check-suse/pull/10


Of the 299 packages with major issues,
there are 3 in Ring0 and 55 in Ring1:

--- ring0 ---
acl
gcc7
gcc8

--- ring1 ---
MozillaFirefox
MozillaThunderbird
apache-commons-lang3
apache-pdfbox
breeze5-icons
bsh2
colord
dolphin
dom4j
ecj
emacs
firebird
gconf2-branding-openSUSE
gdb
gengetopt
gimp
gnome-documents
go1.10
go1.11
groff-full
grub2
hamcrest
installation-images
java-11-openjdk
java-1_8_0-openjdk
javassist
jing-trang
jtidy
junitperf
kernel-debug
kernel-default
kernel-vanilla
kubernetes
libkolabxml
libreoffice
mono-core
mozilla-nss
openmpi2
pcp
php7
plasma5-desktop
python-base
python-numpy
python3-base
release-notes-openSUSE
ruby2.5
rubygem-gem2rpm
rust
syntax-highlighting
tomcat
transfig
wireshark
xmlbeans
xpp2
xpp3


If you are interested in why they vary:
acl:
date+time in /usr/share/locale/en@boldquot/LC_MESSAGES/acl.mo
-PO-Revision-Date: 2018-10-23 01:45+0000
+PO-Revision-Date: 2033-11-24 15:01+0000

gcc7 gcc8:
indeterminism from PGO/parallelism + mtime. See also
https://github.com/bmwiedemann/theunreproduciblepackage/tree/master/pgo


MozillaFirefox MozillaThunderbird:
date+time, other
/usr/lib64/firefox/libxul.so varies maybe from
https://github.com/rust-lang/rust/issues/50556
.../browser/extensions/langpack-ca@xxxxxxxxxxxxxxxxxxx/manifest.json
- "version": "20181009013946"
+ "version": "20331110172641"


apache-commons-lang3 apache-pdfbox dom4j hsqldb javassist jing-trang
jtidy junitperf tomcat xpp2 xpp3:
date + mtime
in .jar + javadoc html


breeze5-icons dolphin wireshark syntax-highlighting:
Qt rcc/qrc fix pending https://codereview.qt-project.org/243636


bsh2:
/usr/share/maven-metadata/bsh2.xml has ASLR order issues
date + mtime in .jar + javadoc html


colord
some randomness in .icc files from cd-create-profile
plus an uncommitted date/time fix in lib/colord/cd-it8.c:
priv->enable_created = FALSE


ecj
ASLR + date?
/usr/share/maven-metadata/ecj.xml has alias entries in random order


emacs
dumps lisp interpreter memory to create its binaries
(similar to clisp)
causes variations in /usr/bin/emacs-*


firebird:
ships unreproducible database
http://tracker.firebirdsql.org/browse/CORE-5548

gconf2-branding-openSUSE
embeds mtime values in
/etc/gconf/gconf.xml.vendor/%gconf-tree.xml

gdb
contains testresults
https://bugzilla.opensuse.org/show_bug.cgi?id=1110708

gimp
ASLR
12 byte differ in bKGD header
from gegl
GEGL_USE_OPENCL=no GEGL_SWAP=ram /usr/bin/gegl
../../icons/Symbolic/64/gimp-texture.png -o 64/gimp-texture.png --
gegl:invert-gamma

gnome-documents
gnome-documents-getting-started.pdf has random ID from inkscape

go1.10 go1.11:
variations in /usr/lib64/go/1.10/pkg/obj/go-build/*/*-a

groff-full
date+time + unknown reason?
.ps files vary
.html files have CreationDate

grub2
mtime + readdir
https://savannah.gnu.org/bugs/index.php?54841

hamcrest
java .class files vary from
build/temp/hamcrest-core/generated-code
/org/hamcrest/CoreMatchers.java written by ant
javadoc html varies

installation-images
many variations from mtimes + filesys/readdir and %post scripts

java-1_8_0-openjdk java-10-openjdk java-11-openjdk
various .jar .zip .html .jmod
ordering in /usr/lib64/jvm/java-10-openjdk-10/lib/classlist

kernel-debug kernel-default kernel-vanilla
date+time ; random keys?
"Build time autogenerated kernel key0
..181009012108Z..21180915012108Z0.1,0"

kubernetes:
fixed upstream: order in man-pages
https://github.com/kubernetes/kubernetes/pull/68983
maybe fixed upstream: random build-ids

libkolabxml
unknown/ASLR?
from build/kolabformat-xcal-schema.cxx
created by compiled/xsdbin.cxx

libreoffice
various .jar .so .dat/.bau (.zip) javadoc_log.txt

mono-core
date+time ; other

mozilla-nss
DSA random temp-key from shlibsign
https://bugzilla.opensuse.org/show_bug.cgi?id=1081723

openmpi2:
date+hostname
proposed https://github.com/open-mpi/ompi/pull/5653 but moving slowly
https://github.com/open-mpi/ompi/issues/3759

pcp
/var/lib/pcp/testsuite/perfevent/perfevent_coverage has random diffs
from gcov / .gcno files causing diff in .o ?

php7
date / EPOCH timestamps
e.g. in /usr/share/php7/PEAR/.channels/__uri.reg

plasma5-desktop
parallelism https://bugzilla.opensuse.org/show_bug.cgi?id=1109420
makes kcolorschemeeditor vary
kcolorschemeeditor_autogen/mocs_compilation.cpp.o differ
workaround:
make -C kcms/colors/

python-base python3-base:
PGO
varies .o files that go into /usr/lib64/libpython3.6m.so.1.0
.pyc files vary
python-base shows influence from date+hostname (in PGO?)

python-numpy
1 .pyc file varied

release-notes-openSUSE
partial fix in https://github.com/openSUSE/daps/issues/482
date+time in .pdf ; needs work on fop java code
random id values in .html

ruby2.5
2 gemspec files have date
created.rid varies

rubygem-gem2rpm
mtime ?
/usr/lib64/ruby/gems/2.5.0/cache/checksums.yaml varies

rust
unknown
asm diffs in cargo and others
- callq
<_ZN4core3ptr13drop_in_place17h2759227fc24294eaE.llvm.10224348644825613051>
+ callq
<_ZN4core3ptr13drop_in_place17h2759227fc24294eaE.llvm.17896929563746027074>


tigervnc
date+time+randomness in
/usr/share/vnc/classes/META-INF/TIGERVNC.RSA
from SignJar.cmake calling jarsigner with temporary private key
partial fix https://github.com/TigerVNC/tigervnc/pull/765

transfig
date+time
in sample-presentation.pdf /CreationDate /ModDate
from pdflatex

virtualbox
tar + random
.so files have random NT_GNU_BUILD_ID (unique build ID bitstring)
maybe from out/linux.amd64/release/obj/webservice/gsoapH_from_gsoap.h
that has date+time in comment
/usr/share/virtualbox/extensions/VNC-5.2.16.vbox-extpack
unhandled gzip content: POSIX tar archive (GNU)
-13343137165.010505. 5...
+16773576625.010527. 5...


xmlbeans
index.xsb varies from filesystem
created in org.apache.xmlbeans.impl.tool.SchemaCompiler


Ciao
Bernhard M.

< Previous Next >
This Thread
  • No further messages