What happened in the Reproducible Builds effort between April 17th and April 23rd 2016:
Toolchain fixes
Thomas Weber uploaded lcms2/2.7-1 which will not write uninitialized memory when writing color names. Original patch by Lunar.
The GCC 7 development phase has just begun, so Dhole reworked his patch to make gcc use SOURCE_DATE_EPOCH if set which prompted interesting feedback, but it has not been merged yet.
Alexis Bienvenüe submitted a patch for sphinx to strip Python object memory addresses from the generated documentation.
Packages fixed
The following packages have become reproducible due to changes in their build dependencies: cobertura, commons-pool, easymock, eclipselink, excalibur-logkit, gap-radiroot, gluegen2, jabref, java3d, jcifs, jline, jmock2, josql, jtharness, libfann, libgroboutils-java, libjemmy2-java, libjgoodies-binding-java, libjgrapht0.8-java, libjtds-java, liboptions-java, libpal-java, libzeus-jscl-java, node-transformers, octave-msh, octave-secs2d, openmama, rkward.
The following packages have become reproducible after being fixed:
- auto-multiple-choice/1.3.0-1 by Alexis Bienvenüe.
- castle-game-engine/5.2.0-3 by Paul Gevers.
- gap-autpgrp/1.5-2 by Bill Allombert.
- gap-grape/4r7+ds-2 by Jerome Benoit.
- letsencrypt.sh/0.1.0-2 by Mattia Rizzolo.
- libindicate/0.6.92-3 by Adam Borowski.
- ltrsift/1.0.2-7 by Sascha Steinbiss.
- perl/5.22.1-10 by Niko Tyni.
- php-pear/1:1.10.1+submodules+notgz-7 by Mathieu Parent.
- python-pint/0.7.2-2 by Ondřej Nový.
- spades/3.7.1+dfsg-3 by Sascha Steinbiss.
- tex4ht/20090611-2 by Reiner Herrmann.
Patches submitted that have not made their way to the archive yet:
- #821356 against emoslib by boyska: use
echoin a portable manner across shells. - #822268 against transdecoder by Dhole: set
PERL_HASH_SEED=0when calling the scripts that generate samples.
tests.reproducible-builds.org
- Steven Chamberlain investigated the performance of our armhf boards which also provided a nice overview of our armhf build network.
- As i386 has almost been completely tested the order of the architectures displayed has been changed to reflect the fact that i386 is now the 2nd most popular architecture in Debian. (h01ger)
- In order to decrease the number of blacklisted packages, the first build is now run with a timeout of 18h (previously: 12h) and the 2nd with 24h timeout (previously: 18h). (h01ger)
- We now also vary the CPU model on amd64 (and soon on i386 too) so that one build is performed using a “AMD Opteron 62xx class CPU” while the other is done using a “Intel Core Processor (Haswell)”. This is now possible as proftitbricks.com offers VMs running both types of CPU and have generously increased their sponsorship once more. (h01ger)
- Profitbricks increased our storage space by 400 GB which will be used to setup a 2nd build node for the coreboot/OpenWrt/NetBSD/Arch Linux/Fedora tests. This 2nd build node will run 398 days in the future for testing reproducibility on a different date.
diffoscope development
diffoscope 52 was released with changes from Mattia Rizzolo, h01ger, Satyam Zode and Reiner Herrmann, who also did the release. Notable changes included:
- Drop transitional debbindiff package.
- Let objdump demangle symbols for better readability.
- Install bin/diffoscope instead of auto-generated script. (Closes: #821777)
As usual, diffoscope 52 is available on Debian, Archlinux and PyPI, other distributions will hopefully soon update.
Package reviews
28 reviews have been added, 11 have been updated and 94 have been removed in this week.
14 FTBFS bugs were reported by Chris Lamb (one being was a duplicate of a bug filed by Sebastian Ramacher an hour earlier).
Misc.
This week’s edition was written by Lunar, Holger ‘h01ger’ Levsen and Chris Lamb and reviewed by a bunch of Reproducible builds folks on IRC.
