Mailinglist Archive: opensuse-factory (602 mails)

< Previous Next >
[opensuse-factory] openSUSE reproducible builds status 2019-02
Hi,

last month's status:
https://lists.opensuse.org/opensuse-factory/2019-01/msg00393.html


Last months' reproducible builds project updates (including my work):
https://reproducible-builds.org/blog/posts/199/
https://reproducible-builds.org/blog/posts/198/
https://reproducible-builds.org/blog/posts/197/
https://reproducible-builds.org/blog/posts/196/


I uploaded https://rb.zq1.de/compare.factory-20190228/ today
and rbstats are:
total-packages: 11405
build-tried: 11405
build-failed: 80
build-n-a: 108
build-succeeded: 11217
build-official-failed+na: 130
build-compare-failed: 287
build-compare-succeeded: 10930
bit-by-bit-identical: 10652
verified-semi-reproducible: 9106
verified-bit-identical: 0
not-bit-by-bit-identical: 569

The new "verified" lines are about comparison of the official build
results with what I got through my local "nachbau" build.

Of the badly unreproducible packages,
5 were in ring0
46 were in ring1

That makes it 51/2744 => 1.86 %
which is below the overall average of
287/11217 => 2.56 %


Fixed core packages since last month:
ant

libqt5-qtscript

pcre2
PGO/parallelism https://build.opensuse.org/request/show/668163

rubygem-gem2rpm

And 4 newly semi-reproducible packages - that means build-compare can
now tell that results are similar enough thanks to
https://github.com/openSUSE/build-compare/pull/31
bsf
jtidy
junit
junitperf


Newly unreproducible core (DVD) packages:
bash
llvm7
apache-commons-codec
bsh2
cglib
eigen3
gnuplot
grep
hsqldb
libqt5-qtquickcontrols
python-pip
python-pycryptodome
sac
scons


If you are interested in why they vary:
apache-commons-codec
html doc variations
autoclassify says: from low-entropy filesystem order non-determinism

bash
PGO-related variations from PID and date and other influences in
profiling run ; since upgrade to 5.0

bsh2 cglib
/usr/share/maven-metadata/bsh2.xml has ASLR order issues

eigen3
latex .log files
=> https://build.opensuse.org/request/show/679669

gnuplot
date+time in eg7.pdf from gnuplot+epstopdf

grep
PGO again

hsqldb
build-compare's javadoc filters do not apply to the files because
their path does not match /usr/*/javadoc/*.html

llvm7
variations in .a files and some debuginfo

libqt5-qtquickcontrols
1-byte-variations in qmlc files from ASLR

python-pip python-pycryptodome
unreproducible .pyc marshalling
https://bugs.python.org/issue34033
low-entropy so may randomly produce same results twice

scons:
hostname => https://github.com/SCons/scons/pull/3312

< Previous Next >
This Thread
  • No further messages